SPB Git

spb/localvm-research Public License

Running LLMs larger than memory on a consumer Mac — falsification-driven research: margin-gated deferred refinement, out-of-core verification on Apple Silicon. TR-01 published.

Python 63.2% JavaScript 23.5% CSS 11.8% Shell 0.9% Makefile 0.5%

expG: full results — margin gating promoted (AUROC 0.85-0.98), affine 2-bit base dead

48 trajectories x 4 bit-widths on Qwen3-1.7B; esc@99%: 60/37/5% at
3/4/8-bit; ParetoQ cliff confirmed at naive 2-bit (3% agreement).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Simon-Pierre Boucher committed 5 h ago (Aug 12, 2026) parent 2b7d23b

Showing 6 changed files with +5,424 and −9

modified experiments/micro/expG_decision_stability/analysis.md +62 −3
@@ -3,10 +3,69 @@ project: localvm-research
3 3 document: expG_decision_stability/analysis
4 4 author: Simon-Pierre Boucher
5 5 contact: contact@spboucher.ai
6 created: 2026-08-11
7 status: draft
6 +created: 2026-08-12
7 +status: reviewed
8 8 ---
9 9
10 10 # Analysis — expG_decision_stability
11 11
12 *To be written after results exist. Must include the seven-field block and the evidence standard of CLAUDE.md §10.*
12 +Run: `results/expG_decision_stability/20260812T042744Z/` · code committed before run.
13 +Model: Qwen3-1.7B (mlx-community bf16 reference), MLX affine quantization
14 +(group 64, all divisible Linear layers) at 2/3/4/8 bits. 48 greedy reference
15 +trajectories (8 per domain × 6 domains), 128 tokens each → 6 144 scored
16 +positions per bit-width. Hardware: M5 Max / 48 GB / macOS 27 (manifest embedded).
17 +
18 +```text
19 +Hypothesis
20 + Margin of the cheap pass predicts disagreement with the full model:
21 + AUROC ≥ 0.8 at 3–4 bits; escalating 10–30% of tokens recovers ≥98–99%.
22 +
23 +Falsification criterion
24 + AUROC < 0.65 at every bit-width, or >50% escalation needed at 4-bit.
25 + → NOT triggered.
26 +
27 +Result (n = 6 144 positions/bit-width)
28 + bits agree mean KL AUROC esc@99% median margin agree/disagree
29 + 2 0.0296 11.89 0.605 99.0% 1.50 / 1.00
30 + 3 0.7310 0.78 0.852 60.2% 5.25 / 1.12
31 + 4 0.8740 0.199 0.898 36.6% 6.50 / 0.75
32 + 8 0.9803 0.0039 0.975 5.1% 6.25 / 0.25
33 + Per-domain AUROC is uniform (4-bit: 0.880–0.921; math best, chat worst);
34 + no domain inverts the signal. Sanity: 4-bit agreement (87.4%) sits near
35 + published ~90–91% 'same top token' rates for 4-bit k-quants on larger
36 + models — plausible for a 1.7B model with plain affine quantization.
37 +
38 +Interpretation
39 + 1. PROMOTE G02 (MarginGate): AUROC 0.852–0.975 clears the 0.8 promote
40 + threshold at 3, 4, and 8 bits. Disagreements concentrate sharply at
41 + small margins (median disagree-margin ≤ 1.1 vs agree-margin ≥ 5.25).
42 + The margin signal is nearly free (it falls out of the cheap forward
43 + pass), and it generalizes across all six domains including reasoning.
44 + 2. G23 (ConfidentlyWrong, contrarian) is weakened but not fully dead:
45 + at 4-bit, reaching 99% agreement still requires escalating 36.6% of
46 + tokens — above the hypothesized 10–30% band, though far below the 50%
47 + kill line. The gate works; whether it is *cheap enough* now depends on
48 + the per-escalation byte cost (expD's question, not this experiment's).
49 + 3. NAIVE 2-BIT AFFINE IS DEAD as a resident base: 3.0% agreement,
50 + KL 11.9, AUROC 0.605 — the model leaves the pretrained basin entirely,
51 + confirming the ParetoQ 2–3-bit cliff on modern SwiGLU models with
52 + amplitude that perplexity tables understate. Candidate C1's resident
53 + base must therefore be ≥3-bit affine, or a codebook/trellis 2-bit
54 + format (which requires the G05 Metal kernel work). This partially
55 + answers G24 (BasinCollapse) for the affine family.
56 + 4. The 8-bit row shows the ceiling: 98% agreement, 5.1% escalation for
57 + 99% — an 8-bit-base system would stream almost nothing but also saves
58 + only 2× memory. The interesting operating region is a 4-bit-class base
59 + (0.5 byte/param resident) escalating ~1/3 of tokens, or a 3-bit base
60 + escalating more; which wins depends on residual fetch granularity.
61 + 5. Note the escalation curves in results.json give the full frontier
62 + (escalated fraction vs residual disagreement) for policy design.
63 +
64 +Next experiment
65 + expD (progressive reconstruction): measure how many *bytes* of residual
66 + (4→8-bit planes, per-layer or per-block) an escalated token actually
67 + needs to flip its decision to the reference — the product
68 + (escalation rate × bytes/escalation) against the expH budget
69 + (~650 MB/token) decides candidate C1's viability. Secondary: rerun expG
70 + at 3B–8B scale to check the expected agreement improvement with size.
71 +```
added experiments/micro/expG_decision_stability/expG_full.log +23 −0
@@ -0,0 +1,23 @@
1 +loading reference mlx-community/Qwen3-1.7B-bf16 …
2 + Fetching 9 files: 0%| | 0/9 [00:00<?, ?it/s] Fetching 9 files: 100%|██████████| 9/9 [00:00<00:00, 5306.26it/s]
3 + generated code
4 + generated math
5 + generated chat
6 + generated french
7 + generated english
8 + generated reasoning
9 +reference generation done in 49s (48 trajectories)
10 +quantizing to 2-bit (group 64) …
11 + Fetching 9 files: 0%| | 0/9 [00:00<?, ?it/s] Fetching 9 files: 100%|██████████| 9/9 [00:00<00:00, 4840.82it/s]
12 + 2-bit: agree=0.0296 AUROC=0.605 esc@99%=0.98974609375
13 +quantizing to 3-bit (group 64) …
14 + Fetching 9 files: 0%| | 0/9 [00:00<?, ?it/s] Fetching 9 files: 100%|██████████| 9/9 [00:00<00:00, 6333.68it/s]
15 + 3-bit: agree=0.7310 AUROC=0.852 esc@99%=0.6017252604166666
16 +quantizing to 4-bit (group 64) …
17 + Fetching 9 files: 0%| | 0/9 [00:00<?, ?it/s] Fetching 9 files: 100%|██████████| 9/9 [00:00<00:00, 4795.32it/s]
18 + 4-bit: agree=0.8740 AUROC=0.898 esc@99%=0.3658854166666667
19 +quantizing to 8-bit (group 64) …
20 + Fetching 9 files: 0%| | 0/9 [00:00<?, ?it/s] Fetching 9 files: 100%|██████████| 9/9 [00:00<00:00, 6190.35it/s]
21 + 8-bit: agree=0.9803 AUROC=0.975 esc@99%=0.051106770833333336
22 +
23 +wrote /Users/simon-pierreboucher/Desktop/localvm-research/results/expG_decision_stability/20260812T042744Z/results.json
modified experiments/micro/expG_decision_stability/hypothesis.md +13 −6
@@ -4,7 +4,8 @@ document: expG_decision_stability/hypothesis
4 4 author: Simon-Pierre Boucher
5 5 contact: contact@spboucher.ai
6 6 created: 2026-08-12
7 status: draft
7 +modified: 2026-08-12
8 +status: reviewed
8 9 ---
9 10
10 11 # Hypothesis — expG_decision_stability
@@ -55,13 +56,19 @@ Baseline
55 56 escalation (loading full weights for all tokens).
56 57
57 58 Result
58 <filled after the run>
59 + CONFIRMED (kill criteria not triggered). AUROC: 0.852 (3-bit), 0.898
60 + (4-bit), 0.975 (8-bit); esc@99%: 60.2% / 36.6% / 5.1%. Naive affine
61 + 2-bit collapsed entirely (3% agreement, AUROC 0.605) — the ParetoQ
62 + cliff, measured. Full numbers: results/expG_decision_stability/
63 + 20260812T042744Z/ and analysis.md.
59 64
60 65 Interpretation
61 <filled after the run>
66 + Margin is a strong, free, domain-uniform gating signal → G02 promoted.
67 + 4-bit escalation need (36.6%) exceeds the hoped 10–30% band, so C1's
68 + economics now hinge on bytes-per-escalation (expD). Resident base must
69 + be ≥3-bit affine or codebook 2-bit (G05).
62 70
63 71 Next experiment
64 <filled after the run; if AUROC high → expD (progressive reconstruction:
65 does escalation *bytes* stay small?); if low → G24 BasinCollapse check
66 before abandoning>
72 + expD — bytes of residual needed per escalated token; then expG rerun at
73 + 3B–8B scale.
67 74 ```
modified research/LOG.md +22 −0
@@ -158,3 +158,25 @@ Format per entry: date/time (local, with timezone) · question · experiment ·
158 158 paper trail: docs, notes, experiments, raw results, code. Content snapshot at commit 9b94283.
159 159 - **Interpretation:** Gate-Zero battery (expG-centric) is the rational first experiment.
160 160 - **Decision:** Next session: implement expG (joint margin×agreement matrix) on a small model.
161 +
162 +---
163 +
164 +## 2026-08-12 10:35 EDT — expG complete: margin-gated escalation PROMOTED, naive 2-bit base DEAD
165 +
166 +- **Question:** Does the cheap pass's top-1 margin predict disagreement with the full model
167 + (the unpublished joint distribution that gates G02/G23/G24 and candidate C1)?
168 +- **Experiment:** expG on Qwen3-1.7B — 48 greedy bf16 trajectories (6 domains), teacher-forced
169 + MLX affine 2/3/4/8-bit variants, 6 144 positions/bit-width.
170 + Results: `results/expG_decision_stability/20260812T042744Z/`.
171 +- **Result:** AUROC(margin→disagreement): 0.852 @3-bit, 0.898 @4-bit, 0.975 @8-bit — all above
172 + the 0.8 promote threshold; per-domain uniform (0.88–0.92 @4-bit). Escalation to reach 99%
173 + agreement: 60.2% / 36.6% / 5.1% of tokens. Median margin agree vs disagree: 6.5 vs 0.75
174 + (4-bit). Naive affine 2-bit collapsed: 3.0% agreement, KL 11.9, AUROC 0.605.
175 +- **Interpretation:** (1) G02 MarginGate PROMOTED — decision uncertainty is a real, free,
176 + domain-stable gating signal. (2) G23 weakened, not dead: 36.6% escalation @4-bit exceeds the
177 + hoped 10–30%; viability now depends on bytes-per-escalation. (3) G24 partially resolved:
178 + the 2–3-bit cliff is real and brutal for affine quantization — C1's resident base must be
179 + ≥3-bit affine or codebook 2-bit (G05 kernel work). (4) 8-bit ceiling: 98% agreement, 5.1%
180 + escalation.
181 +- **Decision:** Implement expD next (residual bytes per escalated token); plan an expG rerun
182 + at 3B–8B to measure the scale trend.
added results/expG_decision_stability/20260812T042628Z/results.json +1095 −0
@@ -0,0 +1,1095 @@
1 +{
2 + "experiment": "expG_decision_stability",
3 + "author": "Simon-Pierre Boucher",
4 + "contact": "contact@spboucher.ai",
5 + "manifest": {
6 + "author": "Simon-Pierre Boucher",
7 + "contact": "contact@spboucher.ai",
8 + "project": "localvm-research",
9 + "collected_utc": "2026-08-12T04:26:28.220460+00:00",
10 + "chip": {
11 + "brand": "Apple M5 Max",
12 + "arch": "arm64",
13 + "cores_total": 18,
14 + "cores_performance": 6,
15 + "cores_efficiency": 12,
16 + "gpu_cores": 40
17 + },
18 + "memory": {
19 + "unified_bytes": 51539607552,
20 + "unified_gb": 48.0,
21 + "pagesize": 16384
22 + },
23 + "ssd": {
24 + "model": "APPLE SSD AP2048Z",
25 + "size": "2 TB",
26 + "smart_status": "Verified"
27 + },
28 + "os": {
29 + "product": "macOS",
30 + "version": "27.0",
31 + "build": "26A5388g",
32 + "kernel": "27.0.0"
33 + },
34 + "software": {
35 + "python": "3.14.4",
36 + "mlx": "0.32.0",
37 + "mlx_lm": "0.31.3",
38 + "torch": null,
39 + "numpy": "2.5.2"
40 + },
41 + "git": {
42 + "commit": "2b7d23b36999ff09c3e6ac03a730575e71e79650",
43 + "dirty_tree": false
44 + },
45 + "thermal_level_at_collect": null
46 + },
47 + "config": {
48 + "model": "mlx-community/Qwen3-1.7B-bf16",
49 + "gen_tokens": 48,
50 + "bits": "4",
51 + "per_domain": 2,
52 + "group_size": 64
53 + },
54 + "n_trajectories": 12,
55 + "results_by_bits": {
56 + "4": {
57 + "bits": 4,
58 + "n_positions": 576,
59 + "agreement_rate": 0.9236111111111112,
60 + "mean_kl_ref_q": 0.11886843293905258,
61 + "auroc_margin_predicts_disagreement": 0.9429895762132604,
62 + "median_margin_agree": 8.8125,
63 + "median_margin_disagree": 0.875,
64 + "escalation_curve": [
65 + {
66 + "tau": 0.0,
67 + "escalated_frac": 0.0,
68 + "residual_disagree": 0.0763888888888889
69 + },
70 + {
71 + "tau": 0.11118090452261309,
72 + "escalated_frac": 0.005208333333333333,
73 + "residual_disagree": 0.07291666666666667
74 + },
75 + {
76 + "tau": 0.22236180904522618,
77 + "escalated_frac": 0.010416666666666666,
78 + "residual_disagree": 0.06944444444444445
79 + },
80 + {
81 + "tau": 0.25,
82 + "escalated_frac": 0.010416666666666666,
83 + "residual_disagree": 0.06944444444444445
84 + },
85 + {
86 + "tau": 0.31972361809045236,
87 + "escalated_frac": 0.020833333333333332,
88 + "residual_disagree": 0.0625
89 + },
90 + {
91 + "tau": 0.375,
92 + "escalated_frac": 0.020833333333333332,
93 + "residual_disagree": 0.0625
94 + },
95 + {
96 + "tau": 0.5,
97 + "escalated_frac": 0.027777777777777776,
98 + "residual_disagree": 0.059027777777777776
99 + },
100 + {
101 + "tau": 0.5,
102 + "escalated_frac": 0.027777777777777776,
103 + "residual_disagree": 0.059027777777777776
104 + },
105 + {
106 + "tau": 0.5,
107 + "escalated_frac": 0.027777777777777776,
108 + "residual_disagree": 0.059027777777777776
109 + },
110 + {
111 + "tau": 0.5,
112 + "escalated_frac": 0.027777777777777776,
113 + "residual_disagree": 0.059027777777777776
114 + },
115 + {
116 + "tau": 0.5,
117 + "escalated_frac": 0.027777777777777776,
118 + "residual_disagree": 0.059027777777777776
119 + },
120 + {
121 + "tau": 0.75,
122 + "escalated_frac": 0.052083333333333336,
123 + "residual_disagree": 0.04513888888888889
124 + },
125 + {
126 + "tau": 0.75,
127 + "escalated_frac": 0.052083333333333336,
128 + "residual_disagree": 0.04513888888888889
129 + },
130 + {
131 + "tau": 0.8203517587939704,
132 + "escalated_frac": 0.06597222222222222,
133 + "residual_disagree": 0.03993055555555555
134 + },
135 + {
136 + "tau": 0.875,
137 + "escalated_frac": 0.06597222222222222,
138 + "residual_disagree": 0.03993055555555555
139 + },
140 + {
141 + "tau": 0.9177135678391961,
142 + "escalated_frac": 0.0763888888888889,
143 + "residual_disagree": 0.034722222222222224
144 + },
145 + {
146 + "tau": 1.0,
147 + "escalated_frac": 0.0763888888888889,
148 + "residual_disagree": 0.034722222222222224
149 + },
150 + {
151 + "tau": 1.0150753768844218,
152 + "escalated_frac": 0.08680555555555555,
153 + "residual_disagree": 0.027777777777777776
154 + },
155 + {
156 + "tau": 1.125,
157 + "escalated_frac": 0.08680555555555555,
158 + "residual_disagree": 0.027777777777777776
159 + },
160 + {
161 + "tau": 1.25,
162 + "escalated_frac": 0.09375,
163 + "residual_disagree": 0.026041666666666668
164 + },
165 + {
166 + "tau": 1.25,
167 + "escalated_frac": 0.09375,
168 + "residual_disagree": 0.026041666666666668
169 + },
170 + {
171 + "tau": 1.25,
172 + "escalated_frac": 0.09375,
173 + "residual_disagree": 0.026041666666666668
174 + },
175 + {
176 + "tau": 1.25,
177 + "escalated_frac": 0.09375,
178 + "residual_disagree": 0.026041666666666668
179 + },
180 + {
181 + "tau": 1.375,
182 + "escalated_frac": 0.11458333333333333,
183 + "residual_disagree": 0.019097222222222224
184 + },
185 + {
186 + "tau": 1.5,
187 + "escalated_frac": 0.11979166666666667,
188 + "residual_disagree": 0.019097222222222224
189 + },
190 + {
191 + "tau": 1.5,
192 + "escalated_frac": 0.11979166666666667,
193 + "residual_disagree": 0.019097222222222224
194 + },
195 + {
196 + "tau": 1.5,
197 + "escalated_frac": 0.11979166666666667,
198 + "residual_disagree": 0.019097222222222224
199 + },
200 + {
201 + "tau": 1.5,
202 + "escalated_frac": 0.11979166666666667,
203 + "residual_disagree": 0.019097222222222224
204 + },
205 + {
206 + "tau": 1.5,
207 + "escalated_frac": 0.11979166666666667,
208 + "residual_disagree": 0.019097222222222224
209 + },
210 + {
211 + "tau": 1.625,
212 + "escalated_frac": 0.1440972222222222,
213 + "residual_disagree": 0.013888888888888888
214 + },
215 + {
216 + "tau": 1.75,
217 + "escalated_frac": 0.14930555555555555,
218 + "residual_disagree": 0.010416666666666666
219 + },
220 + {
221 + "tau": 1.75,
222 + "escalated_frac": 0.14930555555555555,
223 + "residual_disagree": 0.010416666666666666
224 + },
225 + {
226 + "tau": 1.75,
227 + "escalated_frac": 0.14930555555555555,
228 + "residual_disagree": 0.010416666666666666
229 + },
230 + {
231 + "tau": 1.7939698492462313,
232 + "escalated_frac": 0.16666666666666666,
233 + "residual_disagree": 0.010416666666666666
234 + },
235 + {
236 + "tau": 2.0,
237 + "escalated_frac": 0.1684027777777778,
238 + "residual_disagree": 0.010416666666666666
239 + },
240 + {
241 + "tau": 2.0,
242 + "escalated_frac": 0.1684027777777778,
243 + "residual_disagree": 0.010416666666666666
244 + },
245 + {
246 + "tau": 2.125,
247 + "escalated_frac": 0.17881944444444445,
248 + "residual_disagree": 0.010416666666666666
249 + },
250 + {
251 + "tau": 2.25,
252 + "escalated_frac": 0.1840277777777778,
253 + "residual_disagree": 0.010416666666666666
254 + },
255 + {
256 + "tau": 2.25,
257 + "escalated_frac": 0.1840277777777778,
258 + "residual_disagree": 0.010416666666666666
259 + },
260 + {
261 + "tau": 2.25,
262 + "escalated_frac": 0.1840277777777778,
263 + "residual_disagree": 0.010416666666666666
264 + },
265 + {
266 + "tau": 2.3222361809045236,
267 + "escalated_frac": 0.2013888888888889,
268 + "residual_disagree": 0.010416666666666666
269 + },
270 + {
271 + "tau": 2.375,
272 + "escalated_frac": 0.2013888888888889,
273 + "residual_disagree": 0.010416666666666666
274 + },
275 + {
276 + "tau": 2.5,
277 + "escalated_frac": 0.20833333333333334,
278 + "residual_disagree": 0.010416666666666666
279 + },
280 + {
281 + "tau": 2.5,
282 + "escalated_frac": 0.20833333333333334,
283 + "residual_disagree": 0.010416666666666666
284 + },
285 + {
286 + "tau": 2.5,
287 + "escalated_frac": 0.20833333333333334,
288 + "residual_disagree": 0.010416666666666666
289 + },
290 + {
291 + "tau": 2.5031407035175874,
292 + "escalated_frac": 0.22743055555555555,
293 + "residual_disagree": 0.008680555555555556
294 + },
295 + {
296 + "tau": 2.625,
297 + "escalated_frac": 0.22743055555555555,
298 + "residual_disagree": 0.008680555555555556
299 + },
300 + {
301 + "tau": 2.75,
302 + "escalated_frac": 0.2326388888888889,
303 + "residual_disagree": 0.006944444444444444
304 + },
305 + {
306 + "tau": 2.75,
307 + "escalated_frac": 0.2326388888888889,
308 + "residual_disagree": 0.006944444444444444
309 + },
310 + {
311 + "tau": 2.75,
312 + "escalated_frac": 0.2326388888888889,
313 + "residual_disagree": 0.006944444444444444
314 + },
315 + {
316 + "tau": 2.75,
317 + "escalated_frac": 0.2326388888888889,
318 + "residual_disagree": 0.006944444444444444
319 + },
320 + {
321 + "tau": 2.75,
322 + "escalated_frac": 0.2326388888888889,
323 + "residual_disagree": 0.006944444444444444
324 + },
325 + {
326 + "tau": 2.75,
327 + "escalated_frac": 0.2326388888888889,
328 + "residual_disagree": 0.006944444444444444
329 + },
330 + {
331 + "tau": 3.0,
332 + "escalated_frac": 0.265625,
333 + "residual_disagree": 0.003472222222222222
334 + },
335 + {
336 + "tau": 3.0037688442211063,
337 + "escalated_frac": 0.2725694444444444,
338 + "residual_disagree": 0.003472222222222222
339 + },
340 + {
341 + "tau": 3.25,
342 + "escalated_frac": 0.2743055555555556,
343 + "residual_disagree": 0.003472222222222222
344 + },
345 + {
346 + "tau": 3.25,
347 + "escalated_frac": 0.2743055555555556,
348 + "residual_disagree": 0.003472222222222222
349 + },
350 + {
351 + "tau": 3.375,
352 + "escalated_frac": 0.2847222222222222,
353 + "residual_disagree": 0.003472222222222222
354 + },
355 + {
356 + "tau": 3.448492462311556,
357 + "escalated_frac": 0.2916666666666667,
358 + "residual_disagree": 0.003472222222222222
359 + },
360 + {
361 + "tau": 3.5,
362 + "escalated_frac": 0.2916666666666667,
363 + "residual_disagree": 0.003472222222222222
364 + },
365 + {
366 + "tau": 3.5,
367 + "escalated_frac": 0.2916666666666667,
368 + "residual_disagree": 0.003472222222222222
369 + },
370 + {
371 + "tau": 3.75,
372 + "escalated_frac": 0.3055555555555556,
373 + "residual_disagree": 0.003472222222222222
374 + },
375 + {
376 + "tau": 3.75,
377 + "escalated_frac": 0.3055555555555556,
378 + "residual_disagree": 0.003472222222222222
379 + },
380 + {
381 + "tau": 3.8793969849246217,
382 + "escalated_frac": 0.3177083333333333,
383 + "residual_disagree": 0.003472222222222222
384 + },
385 + {
386 + "tau": 4.0,
387 + "escalated_frac": 0.3177083333333333,
388 + "residual_disagree": 0.003472222222222222
389 + },
390 + {
391 + "tau": 4.0,
392 + "escalated_frac": 0.3177083333333333,
393 + "residual_disagree": 0.003472222222222222
394 + },
395 + {
396 + "tau": 4.0,
397 + "escalated_frac": 0.3177083333333333,
398 + "residual_disagree": 0.003472222222222222
399 + },
400 + {
401 + "tau": 4.0,
402 + "escalated_frac": 0.3177083333333333,
403 + "residual_disagree": 0.003472222222222222
404 + },
405 + {
406 + "tau": 4.0,
407 + "escalated_frac": 0.3177083333333333,
408 + "residual_disagree": 0.003472222222222222
409 + },
410 + {
411 + "tau": 4.125,
412 + "escalated_frac": 0.34375,
413 + "residual_disagree": 0.001736111111111111
414 + },
415 + {
416 + "tau": 4.282663316582916,
417 + "escalated_frac": 0.3524305555555556,
418 + "residual_disagree": 0.001736111111111111
419 + },
420 + {
421 + "tau": 4.5,
422 + "escalated_frac": 0.3541666666666667,
423 + "residual_disagree": 0.001736111111111111
424 + },
425 + {
426 + "tau": 4.75,
427 + "escalated_frac": 0.3611111111111111,
428 + "residual_disagree": 0.0
429 + },
430 + {
431 + "tau": 4.75,
432 + "escalated_frac": 0.3611111111111111,
433 + "residual_disagree": 0.0
434 + },
435 + {
436 + "tau": 4.75,
437 + "escalated_frac": 0.3611111111111111,
438 + "residual_disagree": 0.0
439 + },
440 + {
441 + "tau": 4.963567839195981,
442 + "escalated_frac": 0.3767361111111111,
443 + "residual_disagree": 0.0
444 + },
445 + {
446 + "tau": 5.125,
447 + "escalated_frac": 0.3802083333333333,
448 + "residual_disagree": 0.0
449 + },
450 + {
451 + "tau": 5.25,
452 + "escalated_frac": 0.3836805555555556,
453 + "residual_disagree": 0.0
454 + },
455 + {
456 + "tau": 5.25,
457 + "escalated_frac": 0.3836805555555556,
458 + "residual_disagree": 0.0
459 + },
460 + {
461 + "tau": 5.408291457286431,
462 + "escalated_frac": 0.3975694444444444,
463 + "residual_disagree": 0.0
464 + },
465 + {
466 + "tau": 5.5,
467 + "escalated_frac": 0.3975694444444444,
468 + "residual_disagree": 0.0
469 + },
470 + {
471 + "tau": 5.505653266331656,
472 + "escalated_frac": 0.4079861111111111,
473 + "residual_disagree": 0.0
474 + },
475 + {
476 + "tau": 5.625,
477 + "escalated_frac": 0.4079861111111111,
478 + "residual_disagree": 0.0
479 + },
480 + {
481 + "tau": 5.75,
482 + "escalated_frac": 0.4131944444444444,
483 + "residual_disagree": 0.0
484 + },
485 + {
486 + "tau": 5.839195979899497,
487 + "escalated_frac": 0.421875,
488 + "residual_disagree": 0.0
489 + },
490 + {
491 + "tau": 5.950376884422113,
492 + "escalated_frac": 0.4270833333333333,
493 + "residual_disagree": 0.0
494 + },
495 + {
496 + "tau": 6.0,
497 + "escalated_frac": 0.4270833333333333,
498 + "residual_disagree": 0.0
499 + },
500 + {
501 + "tau": 6.172738693467338,
502 + "escalated_frac": 0.4375,
503 + "residual_disagree": 0.0
504 + },
505 + {
506 + "tau": 6.25,
507 + "escalated_frac": 0.4375,
508 + "residual_disagree": 0.0
509 + },
510 + {
511 + "tau": 6.5,
512 + "escalated_frac": 0.4461805555555556,
513 + "residual_disagree": 0.0
514 + },
515 + {
516 + "tau": 6.5,
517 + "escalated_frac": 0.4461805555555556,
518 + "residual_disagree": 0.0
519 + },
520 + {
521 + "tau": 6.5,
522 + "escalated_frac": 0.4461805555555556,
523 + "residual_disagree": 0.0
524 + },
525 + {
526 + "tau": 6.75,
527 + "escalated_frac": 0.4600694444444444,
528 + "residual_disagree": 0.0
529 + },
530 + {
531 + "tau": 7.089824120603019,
532 + "escalated_frac": 0.4670138888888889,
533 + "residual_disagree": 0.0
534 + },
535 + {
536 + "tau": 7.25,
537 + "escalated_frac": 0.4704861111111111,
538 + "residual_disagree": 0.0
539 + },
540 + {
541 + "tau": 7.312185929648244,
542 + "escalated_frac": 0.4774305555555556,
543 + "residual_disagree": 0.0
544 + },
545 + {
546 + "tau": 7.423366834170849,
547 + "escalated_frac": 0.4826388888888889,
548 + "residual_disagree": 0.0
549 + },
550 + {
551 + "tau": 7.5,
552 + "escalated_frac": 0.4826388888888889,
553 + "residual_disagree": 0.0
554 + },
555 + {
556 + "tau": 7.75,
557 + "escalated_frac": 0.4895833333333333,
558 + "residual_disagree": 0.0
559 + },
560 + {
561 + "tau": 7.756909547738694,
562 + "escalated_frac": 0.4982638888888889,
563 + "residual_disagree": 0.0
564 + },
565 + {
566 + "tau": 7.875,
567 + "escalated_frac": 0.4982638888888889,
568 + "residual_disagree": 0.0
569 + },
570 + {
571 + "tau": 8.104271356783919,
572 + "escalated_frac": 0.5069444444444444,
573 + "residual_disagree": 0.0
574 + },
575 + {
576 + "tau": 8.125,
577 + "escalated_frac": 0.5069444444444444,
578 + "residual_disagree": 0.0
579 + },
580 + {
581 + "tau": 8.32663316582915,
582 + "escalated_frac": 0.5173611111111112,
583 + "residual_disagree": 0.0
584 + },
585 + {
586 + "tau": 8.375,
587 + "escalated_frac": 0.5173611111111112,
588 + "residual_disagree": 0.0
589 + },
590 + {
591 + "tau": 8.5,
592 + "escalated_frac": 0.5260416666666666,
593 + "residual_disagree": 0.0
594 + },
595 + {
596 + "tau": 8.75,
597 + "escalated_frac": 0.53125,
598 + "residual_disagree": 0.0
599 + },
600 + {
601 + "tau": 8.7713567839196,
602 + "escalated_frac": 0.5381944444444444,
603 + "residual_disagree": 0.0
604 + },
605 + {
606 + "tau": 9.0,
607 + "escalated_frac": 0.5399305555555556,
608 + "residual_disagree": 0.0
609 + },
610 + {
611 + "tau": 9.0,
612 + "escalated_frac": 0.5399305555555556,
613 + "residual_disagree": 0.0
614 + },
615 + {
616 + "tau": 9.125,
617 + "escalated_frac": 0.5503472222222222,
618 + "residual_disagree": 0.0
619 + },
620 + {
621 + "tau": 9.34108040201005,
622 + "escalated_frac": 0.5572916666666666,
623 + "residual_disagree": 0.0
624 + },
625 + {
626 + "tau": 9.5,
627 + "escalated_frac": 0.5590277777777778,
628 + "residual_disagree": 0.0
629 + },
630 + {
631 + "tau": 9.5,
632 + "escalated_frac": 0.5590277777777778,
633 + "residual_disagree": 0.0
634 + },
635 + {
636 + "tau": 9.625,
637 + "escalated_frac": 0.5694444444444444,
638 + "residual_disagree": 0.0
639 + },
640 + {
641 + "tau": 9.75,
642 + "escalated_frac": 0.5763888888888888,
643 + "residual_disagree": 0.0
644 + },
645 + {
646 + "tau": 9.896984924623112,
647 + "escalated_frac": 0.5833333333333334,
648 + "residual_disagree": 0.0
649 + },
650 + {
651 + "tau": 10.125,
652 + "escalated_frac": 0.5850694444444444,
653 + "residual_disagree": 0.0
654 + },
655 + {
656 + "tau": 10.25,
657 + "escalated_frac": 0.5902777777777778,
658 + "residual_disagree": 0.0
659 + },
660 + {
661 + "tau": 10.375,
662 + "escalated_frac": 0.59375,
663 + "residual_disagree": 0.0
664 + },
665 + {
666 + "tau": 10.5,
667 + "escalated_frac": 0.5989583333333334,
668 + "residual_disagree": 0.0
669 + },
670 + {
671 + "tau": 10.577889447236181,
672 + "escalated_frac": 0.6076388888888888,
673 + "residual_disagree": 0.0
674 + },
675 + {
676 + "tau": 10.625,
677 + "escalated_frac": 0.6076388888888888,
678 + "residual_disagree": 0.0
679 + },
680 + {
681 + "tau": 10.75,
682 + "escalated_frac": 0.6163194444444444,
683 + "residual_disagree": 0.0
684 + },
685 + {
686 + "tau": 10.875,
687 + "escalated_frac": 0.6197916666666666,
688 + "residual_disagree": 0.0
689 + },
690 + {
691 + "tau": 11.0,
692 + "escalated_frac": 0.6267361111111112,
693 + "residual_disagree": 0.0
694 + },
695 + {
696 + "tau": 11.125,
697 + "escalated_frac": 0.6319444444444444,
698 + "residual_disagree": 0.0
699 + },
700 + {
701 + "tau": 11.244974874371856,
702 + "escalated_frac": 0.6371527777777778,
703 + "residual_disagree": 0.0
704 + },
705 + {
706 + "tau": 11.481155778894475,
707 + "escalated_frac": 0.6423611111111112,
708 + "residual_disagree": 0.0
709 + },
710 + {
711 + "tau": 11.684673366834176,
712 + "escalated_frac": 0.6475694444444444,
713 + "residual_disagree": 0.0
714 + },
715 + {
716 + "tau": 11.75,
717 + "escalated_frac": 0.6475694444444444,
718 + "residual_disagree": 0.0
719 + },
720 + {
721 + "tau": 12.0,
722 + "escalated_frac": 0.65625,
723 + "residual_disagree": 0.0
724 + },
725 + {
726 + "tau": 12.25,
727 + "escalated_frac": 0.6614583333333334,
728 + "residual_disagree": 0.0
729 + },
730 + {
731 + "tau": 12.5,
732 + "escalated_frac": 0.6666666666666666,
733 + "residual_disagree": 0.0
734 + },
735 + {
736 + "tau": 12.7964824120603,
737 + "escalated_frac": 0.6736111111111112,
738 + "residual_disagree": 0.0
739 + },
740 + {
741 + "tau": 13.00942211055277,
742 + "escalated_frac": 0.6788194444444444,
743 + "residual_disagree": 0.0
744 + },
745 + {
746 + "tau": 13.370603015075375,
747 + "escalated_frac": 0.6822916666666666,
748 + "residual_disagree": 0.0
749 + },
750 + {
751 + "tau": 13.5,
752 + "escalated_frac": 0.6857638888888888,
753 + "residual_disagree": 0.0
754 + },
755 + {
756 + "tau": 13.625,
757 + "escalated_frac": 0.6892361111111112,
758 + "residual_disagree": 0.0
759 + },
760 + {
761 + "tau": 13.75,
762 + "escalated_frac": 0.6944444444444444,
763 + "residual_disagree": 0.0
764 + },
765 + {
766 + "tau": 13.875,
767 + "escalated_frac": 0.6996527777777778,
768 + "residual_disagree": 0.0
769 + },
770 + {
771 + "tau": 13.875,
772 + "escalated_frac": 0.6996527777777778,
773 + "residual_disagree": 0.0
774 + },
775 + {
776 + "tau": 13.912688442211056,
777 + "escalated_frac": 0.7135416666666666,
778 + "residual_disagree": 0.0
779 + },
780 + {
781 + "tau": 14.148869346733669,
782 + "escalated_frac": 0.71875,
783 + "residual_disagree": 0.0
784 + },
785 + {
786 + "tau": 14.375,
787 + "escalated_frac": 0.7222222222222222,
788 + "residual_disagree": 0.0
789 + },
790 + {
791 + "tau": 14.375,
792 + "escalated_frac": 0.7222222222222222,
793 + "residual_disagree": 0.0
794 + },
795 + {
796 + "tau": 14.375,
797 + "escalated_frac": 0.7222222222222222,
798 + "residual_disagree": 0.0
799 + },
800 + {
801 + "tau": 14.5,
802 + "escalated_frac": 0.734375,
803 + "residual_disagree": 0.0
804 + },
805 + {
806 + "tau": 14.75,
807 + "escalated_frac": 0.7413194444444444,
808 + "residual_disagree": 0.0
809 + },
810 + {
811 + "tau": 14.875,
812 + "escalated_frac": 0.7447916666666666,
813 + "residual_disagree": 0.0
814 + },
815 + {
816 + "tau": 15.0,
817 + "escalated_frac": 0.7517361111111112,
818 + "residual_disagree": 0.0
819 + },
820 + {
821 + "tau": 15.144158291457288,
822 + "escalated_frac": 0.7586805555555556,
823 + "residual_disagree": 0.0
824 + },
825 + {
826 + "tau": 15.375,
827 + "escalated_frac": 0.7621527777777778,
828 + "residual_disagree": 0.0
829 + },
830 + {
831 + "tau": 15.3963567839196,
832 + "escalated_frac": 0.7690972222222222,
833 + "residual_disagree": 0.0
834 + },
835 + {
836 + "tau": 15.746859296482413,
837 + "escalated_frac": 0.7725694444444444,
838 + "residual_disagree": 0.0
839 + },
840 + {
841 + "tau": 15.983040201005025,
842 + "escalated_frac": 0.7777777777777778,
843 + "residual_disagree": 0.0
844 + },
845 + {
846 + "tau": 16.0,
847 + "escalated_frac": 0.7777777777777778,
848 + "residual_disagree": 0.0
849 + },
850 + {
851 + "tau": 16.08040201005025,
852 + "escalated_frac": 0.7881944444444444,
853 + "residual_disagree": 0.0
854 + },
855 + {
856 + "tau": 16.25,
857 + "escalated_frac": 0.7916666666666666,
858 + "residual_disagree": 0.0
859 + },
860 + {
861 + "tau": 16.427763819095482,
862 + "escalated_frac": 0.7986111111111112,
863 + "residual_disagree": 0.0
864 + },
865 + {
866 + "tau": 16.625,
867 + "escalated_frac": 0.8003472222222222,
868 + "residual_disagree": 0.0
869 + },
870 + {
871 + "tau": 16.75,
872 + "escalated_frac": 0.8055555555555556,
873 + "residual_disagree": 0.0
874 + },
875 + {
876 + "tau": 16.886306532663312,
877 + "escalated_frac": 0.8142361111111112,
878 + "residual_disagree": 0.0
879 + },
880 + {
881 + "tau": 17.0,
882 + "escalated_frac": 0.8142361111111112,
883 + "residual_disagree": 0.0
884 + },
885 + {
886 + "tau": 17.125,
887 + "escalated_frac": 0.8211805555555556,
888 + "residual_disagree": 0.0
889 + },
890 + {
891 + "tau": 17.219849246231156,
892 + "escalated_frac": 0.828125,
893 + "residual_disagree": 0.0
894 + },
895 + {
896 + "tau": 17.5,
897 + "escalated_frac": 0.8298611111111112,
898 + "residual_disagree": 0.0
899 + },
900 + {
901 + "tau": 17.625,
902 + "escalated_frac": 0.8368055555555556,
903 + "residual_disagree": 0.0
904 + },
905 + {
906 + "tau": 17.625,
907 + "escalated_frac": 0.8368055555555556,
908 + "residual_disagree": 0.0
909 + },
910 + {
911 + "tau": 17.875,
912 + "escalated_frac": 0.8472222222222222,
913 + "residual_disagree": 0.0
914 + },
915 + {
916 + "tau": 18.05150753768845,
917 + "escalated_frac": 0.8541666666666666,
918 + "residual_disagree": 0.0
919 + },
920 + {
921 + "tau": 18.375,
922 + "escalated_frac": 0.8559027777777778,
923 + "residual_disagree": 0.0
924 + },
925 + {
926 + "tau": 18.5,
927 + "escalated_frac": 0.8611111111111112,
928 + "residual_disagree": 0.0
929 + },
930 + {
931 + "tau": 18.75,
932 + "escalated_frac": 0.8663194444444444,
933 + "residual_disagree": 0.0
934 + },
935 + {
936 + "tau": 18.970477386934675,
937 + "escalated_frac": 0.8732638888888888,
938 + "residual_disagree": 0.0
939 + },
940 + {
941 + "tau": 19.0,
942 + "escalated_frac": 0.8732638888888888,
943 + "residual_disagree": 0.0
944 + },
945 + {
946 + "tau": 19.125,
947 + "escalated_frac": 0.8802083333333334,
948 + "residual_disagree": 0.0
949 + },
950 + {
951 + "tau": 19.179020100502512,
952 + "escalated_frac": 0.8888888888888888,
953 + "residual_disagree": 0.0
954 + },
955 + {
956 + "tau": 19.375,
957 + "escalated_frac": 0.8923611111111112,
958 + "residual_disagree": 0.0
959 + },
960 + {
961 + "tau": 19.526381909547737,
962 + "escalated_frac": 0.8993055555555556,
963 + "residual_disagree": 0.0
964 + },
965 + {
966 + "tau": 19.875,
967 + "escalated_frac": 0.9027777777777778,
968 + "residual_disagree": 0.0
969 + },
970 + {
971 + "tau": 20.123743718592962,
972 + "escalated_frac": 0.9079861111111112,
973 + "residual_disagree": 0.0
974 + },
975 + {
976 + "tau": 20.234924623115575,
977 + "escalated_frac": 0.9131944444444444,
978 + "residual_disagree": 0.0
979 + },
980 + {
981 + "tau": 20.25,
982 + "escalated_frac": 0.9131944444444444,
983 + "residual_disagree": 0.0
984 + },
985 + {
986 + "tau": 20.5822864321608,
987 + "escalated_frac": 0.9236111111111112,
988 + "residual_disagree": 0.0
989 + },
990 + {
991 + "tau": 20.693467336683426,
992 + "escalated_frac": 0.9288194444444444,
993 + "residual_disagree": 0.0
994 + },
995 + {
996 + "tau": 20.80464824120604,
997 + "escalated_frac": 0.9340277777777778,
998 + "residual_disagree": 0.0
999 + },
1000 + {
1001 + "tau": 21.0,
1002 + "escalated_frac": 0.9375,
1003 + "residual_disagree": 0.0
1004 + },
1005 + {
1006 + "tau": 21.081030150753747,
1007 + "escalated_frac": 0.9444444444444444,
1008 + "residual_disagree": 0.0
1009 + },
1010 + {
1011 + "tau": 21.625,
1012 + "escalated_frac": 0.9461805555555556,
1013 + "residual_disagree": 0.0
1014 + },
1015 + {
1016 + "tau": 21.75,
1017 + "escalated_frac": 0.9513888888888888,
1018 + "residual_disagree": 0.0
1019 + },
1020 + {
1021 + "tau": 22.1105527638191,
1022 + "escalated_frac": 0.9583333333333334,
1023 + "residual_disagree": 0.0
1024 + },
1025 + {
1026 + "tau": 22.25,
1027 + "escalated_frac": 0.9618055555555556,
1028 + "residual_disagree": 0.0
1029 + },
1030 + {
1031 + "tau": 23.123743718592976,
1032 + "escalated_frac": 0.96875,
1033 + "residual_disagree": 0.0
1034 + },
1035 + {
1036 + "tau": 23.638190954773876,
1037 + "escalated_frac": 0.9739583333333334,
1038 + "residual_disagree": 0.0
1039 + },
1040 + {
1041 + "tau": 24.858040201005053,
1042 + "escalated_frac": 0.9791666666666666,
1043 + "residual_disagree": 0.0
1044 + },
1045 + {
1046 + "tau": 27.289572864321627,
1047 + "escalated_frac": 0.984375,
1048 + "residual_disagree": 0.0
1049 + },
1050 + {
1051 + "tau": 29.402638190954775,
1052 + "escalated_frac": 0.9895833333333334,
1053 + "residual_disagree": 0.0
1054 + },
1055 + {
1056 + "tau": 31.5,
1057 + "escalated_frac": 0.9930555555555556,
1058 + "residual_disagree": 0.0
1059 + },
1060 + {
1061 + "tau": 31.75,
1062 + "escalated_frac": 0.9982638888888888,
1063 + "residual_disagree": 0.0
1064 + }
1065 + ],
1066 + "per_domain": {
1067 + "code": {
1068 + "agreement_rate": 0.90625,
1069 + "auroc": 0.9086845466155811
1070 + },
1071 + "math": {
1072 + "agreement_rate": 0.9895833333333334,
1073 + "auroc": 0.8842105263157894
1074 + },
1075 + "chat": {
1076 + "agreement_rate": 0.9375,
1077 + "auroc": 0.9462962962962963
1078 + },
1079 + "french": {
1080 + "agreement_rate": 0.8854166666666666,
1081 + "auroc": 0.9368983957219251
1082 + },
1083 + "english": {
1084 + "agreement_rate": 0.8958333333333334,
1085 + "auroc": 0.986046511627907
1086 + },
1087 + "reasoning": {
1088 + "agreement_rate": 0.9270833333333334,
1089 + "auroc": 0.9630818619582665
1090 + }
1091 + },
1092 + "escalation_frac_for_99pct": 0.22743055555555555
1093 + }
1094 + }
1095 +}
\ No newline at end of file
added results/expG_decision_stability/20260812T042744Z/results.json +4209 −0
@@ -0,0 +1,4209 @@
1 +{
2 + "experiment": "expG_decision_stability",
3 + "author": "Simon-Pierre Boucher",
4 + "contact": "contact@spboucher.ai",
5 + "manifest": {
6 + "author": "Simon-Pierre Boucher",
7 + "contact": "contact@spboucher.ai",
8 + "project": "localvm-research",
9 + "collected_utc": "2026-08-12T04:27:44.503861+00:00",
10 + "chip": {
11 + "brand": "Apple M5 Max",
12 + "arch": "arm64",
13 + "cores_total": 18,
14 + "cores_performance": 6,
15 + "cores_efficiency": 12,
16 + "gpu_cores": 40
17 + },
18 + "memory": {
19 + "unified_bytes": 51539607552,
20 + "unified_gb": 48.0,
21 + "pagesize": 16384
22 + },
23 + "ssd": {
24 + "model": "APPLE SSD AP2048Z",
25 + "size": "2 TB",
26 + "smart_status": "Verified"
27 + },
28 + "os": {
29 + "product": "macOS",
30 + "version": "27.0",
31 + "build": "26A5388g",
32 + "kernel": "27.0.0"
33 + },
34 + "software": {
35 + "python": "3.14.4",
36 + "mlx": "0.32.0",
37 + "mlx_lm": "0.31.3",
38 + "torch": null,
39 + "numpy": "2.5.2"
40 + },
41 + "git": {
42 + "commit": "2b7d23b36999ff09c3e6ac03a730575e71e79650",
43 + "dirty_tree": true
44 + },
45 + "thermal_level_at_collect": null
46 + },
47 + "config": {
48 + "model": "mlx-community/Qwen3-1.7B-bf16",
49 + "gen_tokens": 128,
50 + "bits": "2,3,4,8",
51 + "per_domain": 8,
52 + "group_size": 64
53 + },
54 + "n_trajectories": 48,
55 + "results_by_bits": {
56 + "2": {
57 + "bits": 2,
58 + "n_positions": 6144,
59 + "agreement_rate": 0.029622395833333332,
60 + "mean_kl_ref_q": 11.892386436462402,
61 + "auroc_margin_predicts_disagreement": 0.6048301329666643,
62 + "median_margin_agree": 1.5,
63 + "median_margin_disagree": 1.0,
64 + "escalation_curve": [
65 + {
66 + "tau": 0.0,
67 + "escalated_frac": 0.0,
68 + "residual_disagree": 0.9703776041666666
69 + },
70 + {
71 + "tau": 0.0,
72 + "escalated_frac": 0.0,
73 + "residual_disagree": 0.9703776041666666
74 + },
75 + {
76 + "tau": 0.0,
77 + "escalated_frac": 0.0,
78 + "residual_disagree": 0.9703776041666666
79 + },
80 + {
81 + "tau": 0.0,
82 + "escalated_frac": 0.0,
83 + "residual_disagree": 0.9703776041666666
84 + },
85 + {
86 + "tau": 0.0,
87 + "escalated_frac": 0.0,
88 + "residual_disagree": 0.9703776041666666
89 + },
90 + {
91 + "tau": 0.0,
92 + "escalated_frac": 0.0,
93 + "residual_disagree": 0.9703776041666666
94 + },
95 + {
96 + "tau": 0.0,
97 + "escalated_frac": 0.0,
98 + "residual_disagree": 0.9703776041666666
99 + },
100 + {
101 + "tau": 0.0,
102 + "escalated_frac": 0.0,
103 + "residual_disagree": 0.9703776041666666
104 + },
105 + {
106 + "tau": 0.0625,
107 + "escalated_frac": 0.038411458333333336,
108 + "residual_disagree": 0.9327799479166666
109 + },
110 + {
111 + "tau": 0.125,
112 + "escalated_frac": 0.043619791666666664,
113 + "residual_disagree": 0.927734375
114 + },
115 + {
116 + "tau": 0.125,
117 + "escalated_frac": 0.043619791666666664,
118 + "residual_disagree": 0.927734375
119 + },
120 + {
121 + "tau": 0.125,
122 + "escalated_frac": 0.043619791666666664,
123 + "residual_disagree": 0.927734375
124 + },
125 + {
126 + "tau": 0.125,
127 + "escalated_frac": 0.043619791666666664,
128 + "residual_disagree": 0.927734375
129 + },
130 + {
131 + "tau": 0.125,
132 + "escalated_frac": 0.043619791666666664,
133 + "residual_disagree": 0.927734375
134 + },
135 + {
136 + "tau": 0.125,
137 + "escalated_frac": 0.043619791666666664,
138 + "residual_disagree": 0.927734375
139 + },
140 + {
141 + "tau": 0.125,
142 + "escalated_frac": 0.043619791666666664,
143 + "residual_disagree": 0.927734375
144 + },
145 + {
146 + "tau": 0.125,
147 + "escalated_frac": 0.043619791666666664,
148 + "residual_disagree": 0.927734375
149 + },
150 + {
151 + "tau": 0.125,
152 + "escalated_frac": 0.043619791666666664,
153 + "residual_disagree": 0.927734375
154 + },
155 + {
156 + "tau": 0.125,
157 + "escalated_frac": 0.043619791666666664,
158 + "residual_disagree": 0.927734375
159 + },
160 + {
161 + "tau": 0.125,
162 + "escalated_frac": 0.043619791666666664,
163 + "residual_disagree": 0.927734375
164 + },
165 + {
166 + "tau": 0.125,
167 + "escalated_frac": 0.043619791666666664,
168 + "residual_disagree": 0.927734375
169 + },
170 + {
171 + "tau": 0.125,
172 + "escalated_frac": 0.043619791666666664,
173 + "residual_disagree": 0.927734375
174 + },
175 + {
176 + "tau": 0.125,
177 + "escalated_frac": 0.043619791666666664,
178 + "residual_disagree": 0.927734375
179 + },
180 + {
181 + "tau": 0.1875,
182 + "escalated_frac": 0.11328125,
183 + "residual_disagree": 0.8601888020833334
184 + },
185 + {
186 + "tau": 0.25,
187 + "escalated_frac": 0.119140625,
188 + "residual_disagree": 0.8543294270833334
189 + },
190 + {
191 + "tau": 0.25,
192 + "escalated_frac": 0.119140625,
193 + "residual_disagree": 0.8543294270833334
194 + },
195 + {
196 + "tau": 0.25,
197 + "escalated_frac": 0.119140625,
198 + "residual_disagree": 0.8543294270833334
199 + },
200 + {
201 + "tau": 0.25,
202 + "escalated_frac": 0.119140625,
203 + "residual_disagree": 0.8543294270833334
204 + },
205 + {
206 + "tau": 0.25,
207 + "escalated_frac": 0.119140625,
208 + "residual_disagree": 0.8543294270833334
209 + },
210 + {
211 + "tau": 0.25,
212 + "escalated_frac": 0.119140625,
213 + "residual_disagree": 0.8543294270833334
214 + },
215 + {
216 + "tau": 0.25,
217 + "escalated_frac": 0.119140625,
218 + "residual_disagree": 0.8543294270833334
219 + },
220 + {
221 + "tau": 0.25,
222 + "escalated_frac": 0.119140625,
223 + "residual_disagree": 0.8543294270833334
224 + },
225 + {
226 + "tau": 0.25,
227 + "escalated_frac": 0.119140625,
228 + "residual_disagree": 0.8543294270833334
229 + },
230 + {
231 + "tau": 0.25,
232 + "escalated_frac": 0.119140625,
233 + "residual_disagree": 0.8543294270833334
234 + },
235 + {
236 + "tau": 0.25,
237 + "escalated_frac": 0.119140625,
238 + "residual_disagree": 0.8543294270833334
239 + },
240 + {
241 + "tau": 0.25,
242 + "escalated_frac": 0.119140625,
243 + "residual_disagree": 0.8543294270833334
244 + },
245 + {
246 + "tau": 0.25,
247 + "escalated_frac": 0.119140625,
248 + "residual_disagree": 0.8543294270833334
249 + },
250 + {
251 + "tau": 0.25,
252 + "escalated_frac": 0.119140625,
253 + "residual_disagree": 0.8543294270833334
254 + },
255 + {
256 + "tau": 0.3125,
257 + "escalated_frac": 0.18912760416666666,
258 + "residual_disagree": 0.7853190104166666
259 + },
260 + {
261 + "tau": 0.375,
262 + "escalated_frac": 0.19352213541666666,
263 + "residual_disagree": 0.7809244791666666
264 + },
265 + {
266 + "tau": 0.375,
267 + "escalated_frac": 0.19352213541666666,
268 + "residual_disagree": 0.7809244791666666
269 + },
270 + {
271 + "tau": 0.375,
272 + "escalated_frac": 0.19352213541666666,
273 + "residual_disagree": 0.7809244791666666
274 + },
275 + {
276 + "tau": 0.375,
277 + "escalated_frac": 0.19352213541666666,
278 + "residual_disagree": 0.7809244791666666
279 + },
280 + {
281 + "tau": 0.375,
282 + "escalated_frac": 0.19352213541666666,
283 + "residual_disagree": 0.7809244791666666
284 + },
285 + {
286 + "tau": 0.375,
287 + "escalated_frac": 0.19352213541666666,
288 + "residual_disagree": 0.7809244791666666
289 + },
290 + {
291 + "tau": 0.375,
292 + "escalated_frac": 0.19352213541666666,
293 + "residual_disagree": 0.7809244791666666
294 + },
295 + {
296 + "tau": 0.375,
297 + "escalated_frac": 0.19352213541666666,
298 + "residual_disagree": 0.7809244791666666
299 + },
300 + {
301 + "tau": 0.375,
302 + "escalated_frac": 0.19352213541666666,
303 + "residual_disagree": 0.7809244791666666
304 + },
305 + {
306 + "tau": 0.375,
307 + "escalated_frac": 0.19352213541666666,
308 + "residual_disagree": 0.7809244791666666
309 + },
310 + {
311 + "tau": 0.375,
312 + "escalated_frac": 0.19352213541666666,
313 + "residual_disagree": 0.7809244791666666
314 + },
315 + {
316 + "tau": 0.375,
317 + "escalated_frac": 0.19352213541666666,
318 + "residual_disagree": 0.7809244791666666
319 + },
320 + {
321 + "tau": 0.375,
322 + "escalated_frac": 0.19352213541666666,
323 + "residual_disagree": 0.7809244791666666
324 + },
325 + {
326 + "tau": 0.4375,
327 + "escalated_frac": 0.25830078125,
328 + "residual_disagree": 0.7176106770833334
329 + },
330 + {
331 + "tau": 0.5,
332 + "escalated_frac": 0.2630208333333333,
333 + "residual_disagree": 0.712890625
334 + },
335 + {
336 + "tau": 0.5,
337 + "escalated_frac": 0.2630208333333333,
338 + "residual_disagree": 0.712890625
339 + },
340 + {
341 + "tau": 0.5,
342 + "escalated_frac": 0.2630208333333333,
343 + "residual_disagree": 0.712890625
344 + },
345 + {
346 + "tau": 0.5,
347 + "escalated_frac": 0.2630208333333333,
348 + "residual_disagree": 0.712890625
349 + },
350 + {
351 + "tau": 0.5,
352 + "escalated_frac": 0.2630208333333333,
353 + "residual_disagree": 0.712890625
354 + },
355 + {
356 + "tau": 0.5,
357 + "escalated_frac": 0.2630208333333333,
358 + "residual_disagree": 0.712890625
359 + },
360 + {
361 + "tau": 0.5,
362 + "escalated_frac": 0.2630208333333333,
363 + "residual_disagree": 0.712890625
364 + },
365 + {
366 + "tau": 0.5,
367 + "escalated_frac": 0.2630208333333333,
368 + "residual_disagree": 0.712890625
369 + },
370 + {
371 + "tau": 0.5,
372 + "escalated_frac": 0.2630208333333333,
373 + "residual_disagree": 0.712890625
374 + },
375 + {
376 + "tau": 0.5,
377 + "escalated_frac": 0.2630208333333333,
378 + "residual_disagree": 0.712890625
379 + },
380 + {
381 + "tau": 0.5,
382 + "escalated_frac": 0.2630208333333333,
383 + "residual_disagree": 0.712890625
384 + },
385 + {
386 + "tau": 0.5,
387 + "escalated_frac": 0.2630208333333333,
388 + "residual_disagree": 0.712890625
389 + },
390 + {
391 + "tau": 0.5625,
392 + "escalated_frac": 0.3224283854166667,
393 + "residual_disagree": 0.65478515625
394 + },
395 + {
396 + "tau": 0.625,
397 + "escalated_frac": 0.328125,
398 + "residual_disagree": 0.6490885416666666
399 + },
400 + {
401 + "tau": 0.625,
402 + "escalated_frac": 0.328125,
403 + "residual_disagree": 0.6490885416666666
404 + },
405 + {
406 + "tau": 0.625,
407 + "escalated_frac": 0.328125,
408 + "residual_disagree": 0.6490885416666666
409 + },
410 + {
411 + "tau": 0.625,
412 + "escalated_frac": 0.328125,
413 + "residual_disagree": 0.6490885416666666
414 + },
415 + {
416 + "tau": 0.625,
417 + "escalated_frac": 0.328125,
418 + "residual_disagree": 0.6490885416666666
419 + },
420 + {
421 + "tau": 0.625,
422 + "escalated_frac": 0.328125,
423 + "residual_disagree": 0.6490885416666666
424 + },
425 + {
426 + "tau": 0.625,
427 + "escalated_frac": 0.328125,
428 + "residual_disagree": 0.6490885416666666
429 + },
430 + {
431 + "tau": 0.625,
432 + "escalated_frac": 0.328125,
433 + "residual_disagree": 0.6490885416666666
434 + },
435 + {
436 + "tau": 0.625,
437 + "escalated_frac": 0.328125,
438 + "residual_disagree": 0.6490885416666666
439 + },
440 + {
441 + "tau": 0.625,
442 + "escalated_frac": 0.328125,
443 + "residual_disagree": 0.6490885416666666
444 + },
445 + {
446 + "tau": 0.625,
447 + "escalated_frac": 0.328125,
448 + "residual_disagree": 0.6490885416666666
449 + },
450 + {
451 + "tau": 0.75,
452 + "escalated_frac": 0.3841145833333333,
453 + "residual_disagree": 0.5944010416666666
454 + },
455 + {
456 + "tau": 0.75,
457 + "escalated_frac": 0.3841145833333333,
458 + "residual_disagree": 0.5944010416666666
459 + },
460 + {
461 + "tau": 0.75,
462 + "escalated_frac": 0.3841145833333333,
463 + "residual_disagree": 0.5944010416666666
464 + },
465 + {
466 + "tau": 0.75,
467 + "escalated_frac": 0.3841145833333333,
468 + "residual_disagree": 0.5944010416666666
469 + },
470 + {
471 + "tau": 0.75,
472 + "escalated_frac": 0.3841145833333333,
473 + "residual_disagree": 0.5944010416666666
474 + },
475 + {
476 + "tau": 0.75,
477 + "escalated_frac": 0.3841145833333333,
478 + "residual_disagree": 0.5944010416666666
479 + },
480 + {
481 + "tau": 0.75,
482 + "escalated_frac": 0.3841145833333333,
483 + "residual_disagree": 0.5944010416666666
484 + },
485 + {
486 + "tau": 0.75,
487 + "escalated_frac": 0.3841145833333333,
488 + "residual_disagree": 0.5944010416666666
489 + },
490 + {
491 + "tau": 0.75,
492 + "escalated_frac": 0.3841145833333333,
493 + "residual_disagree": 0.5944010416666666
494 + },
495 + {
496 + "tau": 0.75,
497 + "escalated_frac": 0.3841145833333333,
498 + "residual_disagree": 0.5944010416666666
499 + },
500 + {
501 + "tau": 0.8125,
502 + "escalated_frac": 0.4353841145833333,
503 + "residual_disagree": 0.5445963541666666
504 + },
505 + {
506 + "tau": 0.875,
507 + "escalated_frac": 0.4383138020833333,
508 + "residual_disagree": 0.5416666666666666
509 + },
510 + {
511 + "tau": 0.875,
512 + "escalated_frac": 0.4383138020833333,
513 + "residual_disagree": 0.5416666666666666
514 + },
515 + {
516 + "tau": 0.875,
517 + "escalated_frac": 0.4383138020833333,
518 + "residual_disagree": 0.5416666666666666
519 + },
520 + {
521 + "tau": 0.875,
522 + "escalated_frac": 0.4383138020833333,
523 + "residual_disagree": 0.5416666666666666
524 + },
525 + {
526 + "tau": 0.875,
527 + "escalated_frac": 0.4383138020833333,
528 + "residual_disagree": 0.5416666666666666
529 + },
530 + {
531 + "tau": 0.875,
532 + "escalated_frac": 0.4383138020833333,
533 + "residual_disagree": 0.5416666666666666
534 + },
535 + {
536 + "tau": 0.875,
537 + "escalated_frac": 0.4383138020833333,
538 + "residual_disagree": 0.5416666666666666
539 + },
540 + {
541 + "tau": 0.875,
542 + "escalated_frac": 0.4383138020833333,
543 + "residual_disagree": 0.5416666666666666
544 + },
545 + {
546 + "tau": 0.875,
547 + "escalated_frac": 0.4383138020833333,
548 + "residual_disagree": 0.5416666666666666
549 + },
550 + {
551 + "tau": 0.9579145728643255,
552 + "escalated_frac": 0.4874674479166667,
553 + "residual_disagree": 0.4931640625
554 + },
555 + {
556 + "tau": 1.0,
557 + "escalated_frac": 0.4874674479166667,
558 + "residual_disagree": 0.4931640625
559 + },
560 + {
561 + "tau": 1.0,
562 + "escalated_frac": 0.4874674479166667,
563 + "residual_disagree": 0.4931640625
564 + },
565 + {
566 + "tau": 1.0,
567 + "escalated_frac": 0.4874674479166667,
568 + "residual_disagree": 0.4931640625
569 + },
570 + {
571 + "tau": 1.0,
572 + "escalated_frac": 0.4874674479166667,
573 + "residual_disagree": 0.4931640625
574 + },
575 + {
576 + "tau": 1.0,
577 + "escalated_frac": 0.4874674479166667,
578 + "residual_disagree": 0.4931640625
579 + },
580 + {
581 + "tau": 1.0,
582 + "escalated_frac": 0.4874674479166667,
583 + "residual_disagree": 0.4931640625
584 + },
585 + {
586 + "tau": 1.0,
587 + "escalated_frac": 0.4874674479166667,
588 + "residual_disagree": 0.4931640625
589 + },
590 + {
591 + "tau": 1.0,
592 + "escalated_frac": 0.4874674479166667,
593 + "residual_disagree": 0.4931640625
594 + },
595 + {
596 + "tau": 1.125,
597 + "escalated_frac": 0.5319010416666666,
598 + "residual_disagree": 0.4495442708333333
599 + },
600 + {
601 + "tau": 1.125,
602 + "escalated_frac": 0.5319010416666666,
603 + "residual_disagree": 0.4495442708333333
604 + },
605 + {
606 + "tau": 1.125,
607 + "escalated_frac": 0.5319010416666666,
608 + "residual_disagree": 0.4495442708333333
609 + },
610 + {
611 + "tau": 1.125,
612 + "escalated_frac": 0.5319010416666666,
613 + "residual_disagree": 0.4495442708333333
614 + },
615 + {
616 + "tau": 1.125,
617 + "escalated_frac": 0.5319010416666666,
618 + "residual_disagree": 0.4495442708333333
619 + },
620 + {
621 + "tau": 1.125,
622 + "escalated_frac": 0.5319010416666666,
623 + "residual_disagree": 0.4495442708333333
624 + },
625 + {
626 + "tau": 1.125,
627 + "escalated_frac": 0.5319010416666666,
628 + "residual_disagree": 0.4495442708333333
629 + },
630 + {
631 + "tau": 1.1875,
632 + "escalated_frac": 0.56689453125,
633 + "residual_disagree": 0.41552734375
634 + },
635 + {
636 + "tau": 1.25,
637 + "escalated_frac": 0.5686848958333334,
638 + "residual_disagree": 0.4137369791666667
639 + },
640 + {
641 + "tau": 1.25,
642 + "escalated_frac": 0.5686848958333334,
643 + "residual_disagree": 0.4137369791666667
644 + },
645 + {
646 + "tau": 1.25,
647 + "escalated_frac": 0.5686848958333334,
648 + "residual_disagree": 0.4137369791666667
649 + },
650 + {
651 + "tau": 1.25,
652 + "escalated_frac": 0.5686848958333334,
653 + "residual_disagree": 0.4137369791666667
654 + },
655 + {
656 + "tau": 1.25,
657 + "escalated_frac": 0.5686848958333334,
658 + "residual_disagree": 0.4137369791666667
659 + },
660 + {
661 + "tau": 1.25,
662 + "escalated_frac": 0.5686848958333334,
663 + "residual_disagree": 0.4137369791666667
664 + },
665 + {
666 + "tau": 1.375,
667 + "escalated_frac": 0.6022135416666666,
668 + "residual_disagree": 0.3810221354166667
669 + },
670 + {
671 + "tau": 1.375,
672 + "escalated_frac": 0.6022135416666666,
673 + "residual_disagree": 0.3810221354166667
674 + },
675 + {
676 + "tau": 1.375,
677 + "escalated_frac": 0.6022135416666666,
678 + "residual_disagree": 0.3810221354166667
679 + },
680 + {
681 + "tau": 1.375,
682 + "escalated_frac": 0.6022135416666666,
683 + "residual_disagree": 0.3810221354166667
684 + },
685 + {
686 + "tau": 1.375,
687 + "escalated_frac": 0.6022135416666666,
688 + "residual_disagree": 0.3810221354166667
689 + },
690 + {
691 + "tau": 1.375,
692 + "escalated_frac": 0.6022135416666666,
693 + "residual_disagree": 0.3810221354166667
694 + },
695 + {
696 + "tau": 1.4086055276381728,
697 + "escalated_frac": 0.6331380208333334,
698 + "residual_disagree": 0.3518880208333333
699 + },
700 + {
701 + "tau": 1.5,
702 + "escalated_frac": 0.6350911458333334,
703 + "residual_disagree": 0.3499348958333333
704 + },
705 + {
706 + "tau": 1.5,
707 + "escalated_frac": 0.6350911458333334,
708 + "residual_disagree": 0.3499348958333333
709 + },
710 + {
711 + "tau": 1.5,
712 + "escalated_frac": 0.6350911458333334,
713 + "residual_disagree": 0.3499348958333333
714 + },
715 + {
716 + "tau": 1.5,
717 + "escalated_frac": 0.6350911458333334,
718 + "residual_disagree": 0.3499348958333333
719 + },
720 + {
721 + "tau": 1.5,
722 + "escalated_frac": 0.6350911458333334,
723 + "residual_disagree": 0.3499348958333333
724 + },
725 + {
726 + "tau": 1.5625,
727 + "escalated_frac": 0.6629231770833334,
728 + "residual_disagree": 0.3229166666666667
729 + },
730 + {
731 + "tau": 1.625,
732 + "escalated_frac": 0.6640625,
733 + "residual_disagree": 0.32177734375
734 + },
735 + {
736 + "tau": 1.625,
737 + "escalated_frac": 0.6640625,
738 + "residual_disagree": 0.32177734375
739 + },
740 + {
741 + "tau": 1.625,
742 + "escalated_frac": 0.6640625,
743 + "residual_disagree": 0.32177734375
744 + },
745 + {
746 + "tau": 1.625,
747 + "escalated_frac": 0.6640625,
748 + "residual_disagree": 0.32177734375
749 + },
750 + {
751 + "tau": 1.625,
752 + "escalated_frac": 0.6640625,
753 + "residual_disagree": 0.32177734375
754 + },
755 + {
756 + "tau": 1.625,
757 + "escalated_frac": 0.6640625,
758 + "residual_disagree": 0.32177734375
759 + },
760 + {
761 + "tau": 1.75,
762 + "escalated_frac": 0.697265625,
763 + "residual_disagree": 0.2898763020833333
764 + },
765 + {
766 + "tau": 1.75,
767 + "escalated_frac": 0.697265625,
768 + "residual_disagree": 0.2898763020833333
769 + },
770 + {
771 + "tau": 1.75,
772 + "escalated_frac": 0.697265625,
773 + "residual_disagree": 0.2898763020833333
774 + },
775 + {
776 + "tau": 1.75,
777 + "escalated_frac": 0.697265625,
778 + "residual_disagree": 0.2898763020833333
779 + },
780 + {
781 + "tau": 1.75,
782 + "escalated_frac": 0.697265625,
783 + "residual_disagree": 0.2898763020833333
784 + },
785 + {
786 + "tau": 1.875,
787 + "escalated_frac": 0.7234700520833334,
788 + "residual_disagree": 0.2646484375
789 + },
790 + {
791 + "tau": 1.875,
792 + "escalated_frac": 0.7234700520833334,
793 + "residual_disagree": 0.2646484375
794 + },
795 + {
796 + "tau": 1.875,
797 + "escalated_frac": 0.7234700520833334,
798 + "residual_disagree": 0.2646484375
799 + },
800 + {
801 + "tau": 1.875,
802 + "escalated_frac": 0.7234700520833334,
803 + "residual_disagree": 0.2646484375
804 + },
805 + {
806 + "tau": 1.875,
807 + "escalated_frac": 0.7234700520833334,
808 + "residual_disagree": 0.2646484375
809 + },
810 + {
811 + "tau": 2.0,
812 + "escalated_frac": 0.7451171875,
813 + "residual_disagree": 0.24348958333333334
814 + },
815 + {
816 + "tau": 2.0,
817 + "escalated_frac": 0.7451171875,
818 + "residual_disagree": 0.24348958333333334
819 + },
820 + {
821 + "tau": 2.0,
822 + "escalated_frac": 0.7451171875,
823 + "residual_disagree": 0.24348958333333334
824 + },
825 + {
826 + "tau": 2.0,
827 + "escalated_frac": 0.7451171875,
828 + "residual_disagree": 0.24348958333333334
829 + },
830 + {
831 + "tau": 2.125,
832 + "escalated_frac": 0.76708984375,
833 + "residual_disagree": 0.22184244791666666
834 + },
835 + {
836 + "tau": 2.125,
837 + "escalated_frac": 0.76708984375,
838 + "residual_disagree": 0.22184244791666666
839 + },
840 + {
841 + "tau": 2.125,
842 + "escalated_frac": 0.76708984375,
843 + "residual_disagree": 0.22184244791666666
844 + },
845 + {
846 + "tau": 2.125,
847 + "escalated_frac": 0.76708984375,
848 + "residual_disagree": 0.22184244791666666
849 + },
850 + {
851 + "tau": 2.25,
852 + "escalated_frac": 0.787109375,
853 + "residual_disagree": 0.20263671875
854 + },
855 + {
856 + "tau": 2.25,
857 + "escalated_frac": 0.787109375,
858 + "residual_disagree": 0.20263671875
859 + },
860 + {
861 + "tau": 2.25,
862 + "escalated_frac": 0.787109375,
863 + "residual_disagree": 0.20263671875
864 + },
865 + {
866 + "tau": 2.375,
867 + "escalated_frac": 0.8030598958333334,
868 + "residual_disagree": 0.18701171875
869 + },
870 + {
871 + "tau": 2.375,
872 + "escalated_frac": 0.8030598958333334,
873 + "residual_disagree": 0.18701171875
874 + },
875 + {
876 + "tau": 2.375,
877 + "escalated_frac": 0.8030598958333334,
878 + "residual_disagree": 0.18701171875
879 + },
880 + {
881 + "tau": 2.375,
882 + "escalated_frac": 0.8030598958333334,
883 + "residual_disagree": 0.18701171875
884 + },
885 + {
886 + "tau": 2.5,
887 + "escalated_frac": 0.8204752604166666,
888 + "residual_disagree": 0.169921875
889 + },
890 + {
891 + "tau": 2.5,
892 + "escalated_frac": 0.8204752604166666,
893 + "residual_disagree": 0.169921875
894 + },
895 + {
896 + "tau": 2.5,
897 + "escalated_frac": 0.8204752604166666,
898 + "residual_disagree": 0.169921875
899 + },
900 + {
901 + "tau": 2.625,
902 + "escalated_frac": 0.8385416666666666,
903 + "residual_disagree": 0.15234375
904 + },
905 + {
906 + "tau": 2.625,
907 + "escalated_frac": 0.8385416666666666,
908 + "residual_disagree": 0.15234375
909 + },
910 + {
911 + "tau": 2.625,
912 + "escalated_frac": 0.8385416666666666,
913 + "residual_disagree": 0.15234375
914 + },
915 + {
916 + "tau": 2.75,
917 + "escalated_frac": 0.853515625,
918 + "residual_disagree": 0.13736979166666666
919 + },
920 + {
921 + "tau": 2.75,
922 + "escalated_frac": 0.853515625,
923 + "residual_disagree": 0.13736979166666666
924 + },
925 + {
926 + "tau": 2.75,
927 + "escalated_frac": 0.853515625,
928 + "residual_disagree": 0.13736979166666666
929 + },
930 + {
931 + "tau": 2.875,
932 + "escalated_frac": 0.86669921875,
933 + "residual_disagree": 0.12483723958333333
934 + },
935 + {
936 + "tau": 2.875,
937 + "escalated_frac": 0.86669921875,
938 + "residual_disagree": 0.12483723958333333
939 + },
940 + {
941 + "tau": 2.875,
942 + "escalated_frac": 0.86669921875,
943 + "residual_disagree": 0.12483723958333333
944 + },
945 + {
946 + "tau": 3.0,
947 + "escalated_frac": 0.88134765625,
948 + "residual_disagree": 0.11051432291666667
949 + },
950 + {
951 + "tau": 3.0,
952 + "escalated_frac": 0.88134765625,
953 + "residual_disagree": 0.11051432291666667
954 + },
955 + {
956 + "tau": 3.125,
957 + "escalated_frac": 0.8942057291666666,
958 + "residual_disagree": 0.09814453125
959 + },
960 + {
961 + "tau": 3.125,
962 + "escalated_frac": 0.8942057291666666,
963 + "residual_disagree": 0.09814453125
964 + },
965 + {
966 + "tau": 3.217650753768851,
967 + "escalated_frac": 0.9044596354166666,
968 + "residual_disagree": 0.08837890625
969 + },
970 + {
971 + "tau": 3.25,
972 + "escalated_frac": 0.9044596354166666,
973 + "residual_disagree": 0.08837890625
974 + },
975 + {
976 + "tau": 3.375,
977 + "escalated_frac": 0.9130859375,
978 + "residual_disagree": 0.08024088541666667
979 + },
980 + {
981 + "tau": 3.375,
982 + "escalated_frac": 0.9130859375,
983 + "residual_disagree": 0.08024088541666667
984 + },
985 + {
986 + "tau": 3.5,
987 + "escalated_frac": 0.92041015625,
988 + "residual_disagree": 0.07307942708333333
989 + },
990 + {
991 + "tau": 3.625,
992 + "escalated_frac": 0.9267578125,
993 + "residual_disagree": 0.06722005208333333
994 + },
995 + {
996 + "tau": 3.75,
997 + "escalated_frac": 0.93310546875,
998 + "residual_disagree": 0.0615234375
999 + },
1000 + {
1001 + "tau": 3.75,
1002 + "escalated_frac": 0.93310546875,
1003 + "residual_disagree": 0.0615234375
1004 + },
1005 + {
1006 + "tau": 3.875,
1007 + "escalated_frac": 0.9401041666666666,
1008 + "residual_disagree": 0.054850260416666664
1009 + },
1010 + {
1011 + "tau": 4.0,
1012 + "escalated_frac": 0.9456380208333334,
1013 + "residual_disagree": 0.049641927083333336
1014 + },
1015 + {
1016 + "tau": 4.125,
1017 + "escalated_frac": 0.95166015625,
1018 + "residual_disagree": 0.0439453125
1019 + },
1020 + {
1021 + "tau": 4.25,
1022 + "escalated_frac": 0.9560546875,
1023 + "residual_disagree": 0.039876302083333336
1024 + },
1025 + {
1026 + "tau": 4.5,
1027 + "escalated_frac": 0.9637044270833334,
1028 + "residual_disagree": 0.032552083333333336
1029 + },
1030 + {
1031 + "tau": 4.75,
1032 + "escalated_frac": 0.9690755208333334,
1033 + "residual_disagree": 0.02734375
1034 + },
1035 + {
1036 + "tau": 5.0,
1037 + "escalated_frac": 0.974609375,
1038 + "residual_disagree": 0.022298177083333332
1039 + },
1040 + {
1041 + "tau": 5.25,
1042 + "escalated_frac": 0.97802734375,
1043 + "residual_disagree": 0.01904296875
1044 + },
1045 + {
1046 + "tau": 5.75,
1047 + "escalated_frac": 0.9847005208333334,
1048 + "residual_disagree": 0.01318359375
1049 + },
1050 + {
1051 + "tau": 6.25,
1052 + "escalated_frac": 0.98974609375,
1053 + "residual_disagree": 0.008463541666666666
1054 + },
1055 + {
1056 + "tau": 7.25,
1057 + "escalated_frac": 0.9947916666666666,
1058 + "residual_disagree": 0.004069010416666667
1059 + },
1060 + {
1061 + "tau": 14.875,
1062 + "escalated_frac": 0.9998372395833334,
1063 + "residual_disagree": 0.00016276041666666666
1064 + }
1065 + ],
1066 + "per_domain": {
1067 + "code": {
1068 + "agreement_rate": 0.017578125,
1069 + "auroc": 0.5606914071128782
1070 + },
1071 + "math": {
1072 + "agreement_rate": 0.0546875,
1073 + "auroc": 0.6518410566706021
1074 + },
1075 + "chat": {
1076 + "agreement_rate": 0.0205078125,
1077 + "auroc": 0.5552153064615677
1078 + },
1079 + "french": {
1080 + "agreement_rate": 0.0263671875,
1081 + "auroc": 0.6100338051190609
1082 + },
1083 + "english": {
1084 + "agreement_rate": 0.029296875,
1085 + "auroc": 0.6428571428571429
1086 + },
1087 + "reasoning": {
1088 + "agreement_rate": 0.029296875,
1089 + "auroc": 0.5358484238765929
1090 + }
1091 + },
1092 + "escalation_frac_for_99pct": 0.98974609375
1093 + },
1094 + "3": {
1095 + "bits": 3,
1096 + "n_positions": 6144,
1097 + "agreement_rate": 0.73095703125,
1098 + "mean_kl_ref_q": 0.7811363339424133,
1099 + "auroc_margin_predicts_disagreement": 0.8521382349292252,
1100 + "median_margin_agree": 5.25,
1101 + "median_margin_disagree": 1.125,
1102 + "escalation_curve": [
1103 + {
1104 + "tau": 0.0,
1105 + "escalated_frac": 0.0,
1106 + "residual_disagree": 0.26904296875
1107 + },
1108 + {
1109 + "tau": 0.0,
1110 + "escalated_frac": 0.0,
1111 + "residual_disagree": 0.26904296875
1112 + },
1113 + {
1114 + "tau": 0.0,
1115 + "escalated_frac": 0.0,
1116 + "residual_disagree": 0.26904296875
1117 + },
1118 + {
1119 + "tau": 0.0,
1120 + "escalated_frac": 0.0,
1121 + "residual_disagree": 0.26904296875
1122 + },
1123 + {
1124 + "tau": 0.0,
1125 + "escalated_frac": 0.0,
1126 + "residual_disagree": 0.26904296875
1127 + },
1128 + {
1129 + "tau": 0.125,
1130 + "escalated_frac": 0.020833333333333332,
1131 + "residual_disagree": 0.2550455729166667
1132 + },
1133 + {
1134 + "tau": 0.125,
1135 + "escalated_frac": 0.020833333333333332,
1136 + "residual_disagree": 0.2550455729166667
1137 + },
1138 + {
1139 + "tau": 0.125,
1140 + "escalated_frac": 0.020833333333333332,
1141 + "residual_disagree": 0.2550455729166667
1142 + },
1143 + {
1144 + "tau": 0.25,
1145 + "escalated_frac": 0.038248697916666664,
1146 + "residual_disagree": 0.2421875
1147 + },
1148 + {
1149 + "tau": 0.25,
1150 + "escalated_frac": 0.038248697916666664,
1151 + "residual_disagree": 0.2421875
1152 + },
1153 + {
1154 + "tau": 0.25,
1155 + "escalated_frac": 0.038248697916666664,
1156 + "residual_disagree": 0.2421875
1157 + },
1158 + {
1159 + "tau": 0.25,
1160 + "escalated_frac": 0.038248697916666664,
1161 + "residual_disagree": 0.2421875
1162 + },
1163 + {
1164 + "tau": 0.25,
1165 + "escalated_frac": 0.038248697916666664,
1166 + "residual_disagree": 0.2421875
1167 + },
1168 + {
1169 + "tau": 0.25,
1170 + "escalated_frac": 0.038248697916666664,
1171 + "residual_disagree": 0.2421875
1172 + },
1173 + {
1174 + "tau": 0.25,
1175 + "escalated_frac": 0.038248697916666664,
1176 + "residual_disagree": 0.2421875
1177 + },
1178 + {
1179 + "tau": 0.375,
1180 + "escalated_frac": 0.07096354166666667,
1181 + "residual_disagree": 0.21988932291666666
1182 + },
1183 + {
1184 + "tau": 0.375,
1185 + "escalated_frac": 0.07096354166666667,
1186 + "residual_disagree": 0.21988932291666666
1187 + },
1188 + {
1189 + "tau": 0.375,
1190 + "escalated_frac": 0.07096354166666667,
1191 + "residual_disagree": 0.21988932291666666
1192 + },
1193 + {
1194 + "tau": 0.5,
1195 + "escalated_frac": 0.08723958333333333,
1196 + "residual_disagree": 0.20930989583333334
1197 + },
1198 + {
1199 + "tau": 0.5,
1200 + "escalated_frac": 0.08723958333333333,
1201 + "residual_disagree": 0.20930989583333334
1202 + },
1203 + {
1204 + "tau": 0.5,
1205 + "escalated_frac": 0.08723958333333333,
1206 + "residual_disagree": 0.20930989583333334
1207 + },
1208 + {
1209 + "tau": 0.5,
1210 + "escalated_frac": 0.08723958333333333,
1211 + "residual_disagree": 0.20930989583333334
1212 + },
1213 + {
1214 + "tau": 0.5,
1215 + "escalated_frac": 0.08723958333333333,
1216 + "residual_disagree": 0.20930989583333334
1217 + },
1218 + {
1219 + "tau": 0.5,
1220 + "escalated_frac": 0.08723958333333333,
1221 + "residual_disagree": 0.20930989583333334
1222 + },
1223 + {
1224 + "tau": 0.625,
1225 + "escalated_frac": 0.11946614583333333,
1226 + "residual_disagree": 0.1904296875
1227 + },
1228 + {
1229 + "tau": 0.625,
1230 + "escalated_frac": 0.11946614583333333,
1231 + "residual_disagree": 0.1904296875
1232 + },
1233 + {
1234 + "tau": 0.625,
1235 + "escalated_frac": 0.11946614583333333,
1236 + "residual_disagree": 0.1904296875
1237 + },
1238 + {
1239 + "tau": 0.625,
1240 + "escalated_frac": 0.11946614583333333,
1241 + "residual_disagree": 0.1904296875
1242 + },
1243 + {
1244 + "tau": 0.75,
1245 + "escalated_frac": 0.13688151041666666,
1246 + "residual_disagree": 0.18001302083333334
1247 + },
1248 + {
1249 + "tau": 0.75,
1250 + "escalated_frac": 0.13688151041666666,
1251 + "residual_disagree": 0.18001302083333334
1252 + },
1253 + {
1254 + "tau": 0.75,
1255 + "escalated_frac": 0.13688151041666666,
1256 + "residual_disagree": 0.18001302083333334
1257 + },
1258 + {
1259 + "tau": 0.75,
1260 + "escalated_frac": 0.13688151041666666,
1261 + "residual_disagree": 0.18001302083333334
1262 + },
1263 + {
1264 + "tau": 0.75,
1265 + "escalated_frac": 0.13688151041666666,
1266 + "residual_disagree": 0.18001302083333334
1267 + },
1268 + {
1269 + "tau": 0.875,
1270 + "escalated_frac": 0.16552734375,
1271 + "residual_disagree": 0.16162109375
1272 + },
1273 + {
1274 + "tau": 0.875,
1275 + "escalated_frac": 0.16552734375,
1276 + "residual_disagree": 0.16162109375
1277 + },
1278 + {
1279 + "tau": 0.875,
1280 + "escalated_frac": 0.16552734375,
1281 + "residual_disagree": 0.16162109375
1282 + },
1283 + {
1284 + "tau": 0.875,
1285 + "escalated_frac": 0.16552734375,
1286 + "residual_disagree": 0.16162109375
1287 + },
1288 + {
1289 + "tau": 1.0,
1290 + "escalated_frac": 0.18131510416666666,
1291 + "residual_disagree": 0.15185546875
1292 + },
1293 + {
1294 + "tau": 1.0,
1295 + "escalated_frac": 0.18131510416666666,
1296 + "residual_disagree": 0.15185546875
1297 + },
1298 + {
1299 + "tau": 1.0,
1300 + "escalated_frac": 0.18131510416666666,
1301 + "residual_disagree": 0.15185546875
1302 + },
1303 + {
1304 + "tau": 1.0,
1305 + "escalated_frac": 0.18131510416666666,
1306 + "residual_disagree": 0.15185546875
1307 + },
1308 + {
1309 + "tau": 1.0,
1310 + "escalated_frac": 0.18131510416666666,
1311 + "residual_disagree": 0.15185546875
1312 + },
1313 + {
1314 + "tau": 1.0,
1315 + "escalated_frac": 0.18131510416666666,
1316 + "residual_disagree": 0.15185546875
1317 + },
1318 + {
1319 + "tau": 1.125,
1320 + "escalated_frac": 0.21223958333333334,
1321 + "residual_disagree": 0.13525390625
1322 + },
1323 + {
1324 + "tau": 1.125,
1325 + "escalated_frac": 0.21223958333333334,
1326 + "residual_disagree": 0.13525390625
1327 + },
1328 + {
1329 + "tau": 1.25,
1330 + "escalated_frac": 0.22574869791666666,
1331 + "residual_disagree": 0.12890625
1332 + },
1333 + {
1334 + "tau": 1.25,
1335 + "escalated_frac": 0.22574869791666666,
1336 + "residual_disagree": 0.12890625
1337 + },
1338 + {
1339 + "tau": 1.25,
1340 + "escalated_frac": 0.22574869791666666,
1341 + "residual_disagree": 0.12890625
1342 + },
1343 + {
1344 + "tau": 1.25,
1345 + "escalated_frac": 0.22574869791666666,
1346 + "residual_disagree": 0.12890625
1347 + },
1348 + {
1349 + "tau": 1.25,
1350 + "escalated_frac": 0.22574869791666666,
1351 + "residual_disagree": 0.12890625
1352 + },
1353 + {
1354 + "tau": 1.25,
1355 + "escalated_frac": 0.22574869791666666,
1356 + "residual_disagree": 0.12890625
1357 + },
1358 + {
1359 + "tau": 1.375,
1360 + "escalated_frac": 0.2521158854166667,
1361 + "residual_disagree": 0.11555989583333333
1362 + },
1363 + {
1364 + "tau": 1.375,
1365 + "escalated_frac": 0.2521158854166667,
1366 + "residual_disagree": 0.11555989583333333
1367 + },
1368 + {
1369 + "tau": 1.5,
1370 + "escalated_frac": 0.2635091145833333,
1371 + "residual_disagree": 0.10807291666666667
1372 + },
1373 + {
1374 + "tau": 1.5,
1375 + "escalated_frac": 0.2635091145833333,
1376 + "residual_disagree": 0.10807291666666667
1377 + },
1378 + {
1379 + "tau": 1.5,
1380 + "escalated_frac": 0.2635091145833333,
1381 + "residual_disagree": 0.10807291666666667
1382 + },
1383 + {
1384 + "tau": 1.5,
1385 + "escalated_frac": 0.2635091145833333,
1386 + "residual_disagree": 0.10807291666666667
1387 + },
1388 + {
1389 + "tau": 1.625,
1390 + "escalated_frac": 0.28466796875,
1391 + "residual_disagree": 0.09733072916666667
1392 + },
1393 + {
1394 + "tau": 1.625,
1395 + "escalated_frac": 0.28466796875,
1396 + "residual_disagree": 0.09733072916666667
1397 + },
1398 + {
1399 + "tau": 1.625,
1400 + "escalated_frac": 0.28466796875,
1401 + "residual_disagree": 0.09733072916666667
1402 + },
1403 + {
1404 + "tau": 1.75,
1405 + "escalated_frac": 0.2985026041666667,
1406 + "residual_disagree": 0.09000651041666667
1407 + },
1408 + {
1409 + "tau": 1.75,
1410 + "escalated_frac": 0.2985026041666667,
1411 + "residual_disagree": 0.09000651041666667
1412 + },
1413 + {
1414 + "tau": 1.75,
1415 + "escalated_frac": 0.2985026041666667,
1416 + "residual_disagree": 0.09000651041666667
1417 + },
1418 + {
1419 + "tau": 1.75,
1420 + "escalated_frac": 0.2985026041666667,
1421 + "residual_disagree": 0.09000651041666667
1422 + },
1423 + {
1424 + "tau": 1.75,
1425 + "escalated_frac": 0.2985026041666667,
1426 + "residual_disagree": 0.09000651041666667
1427 + },
1428 + {
1429 + "tau": 1.875,
1430 + "escalated_frac": 0.32177734375,
1431 + "residual_disagree": 0.07926432291666667
1432 + },
1433 + {
1434 + "tau": 1.875,
1435 + "escalated_frac": 0.32177734375,
1436 + "residual_disagree": 0.07926432291666667
1437 + },
1438 + {
1439 + "tau": 2.0,
1440 + "escalated_frac": 0.33203125,
1441 + "residual_disagree": 0.0751953125
1442 + },
1443 + {
1444 + "tau": 2.0,
1445 + "escalated_frac": 0.33203125,
1446 + "residual_disagree": 0.0751953125
1447 + },
1448 + {
1449 + "tau": 2.0,
1450 + "escalated_frac": 0.33203125,
1451 + "residual_disagree": 0.0751953125
1452 + },
1453 + {
1454 + "tau": 2.0,
1455 + "escalated_frac": 0.33203125,
1456 + "residual_disagree": 0.0751953125
1457 + },
1458 + {
1459 + "tau": 2.125,
1460 + "escalated_frac": 0.3527018229166667,
1461 + "residual_disagree": 0.068359375
1462 + },
1463 + {
1464 + "tau": 2.125,
1465 + "escalated_frac": 0.3527018229166667,
1466 + "residual_disagree": 0.068359375
1467 + },
1468 + {
1469 + "tau": 2.25,
1470 + "escalated_frac": 0.3619791666666667,
1471 + "residual_disagree": 0.06461588541666667
1472 + },
1473 + {
1474 + "tau": 2.25,
1475 + "escalated_frac": 0.3619791666666667,
1476 + "residual_disagree": 0.06461588541666667
1477 + },
1478 + {
1479 + "tau": 2.25,
1480 + "escalated_frac": 0.3619791666666667,
1481 + "residual_disagree": 0.06461588541666667
1482 + },
1483 + {
1484 + "tau": 2.375,
1485 + "escalated_frac": 0.3811848958333333,
1486 + "residual_disagree": 0.057942708333333336
1487 + },
1488 + {
1489 + "tau": 2.375,
1490 + "escalated_frac": 0.3811848958333333,
1491 + "residual_disagree": 0.057942708333333336
1492 + },
1493 + {
1494 + "tau": 2.5,
1495 + "escalated_frac": 0.3907877604166667,
1496 + "residual_disagree": 0.0546875
1497 + },
1498 + {
1499 + "tau": 2.5,
1500 + "escalated_frac": 0.3907877604166667,
1501 + "residual_disagree": 0.0546875
1502 + },
1503 + {
1504 + "tau": 2.5,
1505 + "escalated_frac": 0.3907877604166667,
1506 + "residual_disagree": 0.0546875
1507 + },
1508 + {
1509 + "tau": 2.5,
1510 + "escalated_frac": 0.3907877604166667,
1511 + "residual_disagree": 0.0546875
1512 + },
1513 + {
1514 + "tau": 2.625,
1515 + "escalated_frac": 0.4083658854166667,
1516 + "residual_disagree": 0.04931640625
1517 + },
1518 + {
1519 + "tau": 2.75,
1520 + "escalated_frac": 0.41650390625,
1521 + "residual_disagree": 0.046549479166666664
1522 + },
1523 + {
1524 + "tau": 2.75,
1525 + "escalated_frac": 0.41650390625,
1526 + "residual_disagree": 0.046549479166666664
1527 + },
1528 + {
1529 + "tau": 2.75,
1530 + "escalated_frac": 0.41650390625,
1531 + "residual_disagree": 0.046549479166666664
1532 + },
1533 + {
1534 + "tau": 2.75,
1535 + "escalated_frac": 0.41650390625,
1536 + "residual_disagree": 0.046549479166666664
1537 + },
1538 + {
1539 + "tau": 2.875,
1540 + "escalated_frac": 0.4334309895833333,
1541 + "residual_disagree": 0.040201822916666664
1542 + },
1543 + {
1544 + "tau": 3.0,
1545 + "escalated_frac": 0.4404296875,
1546 + "residual_disagree": 0.038248697916666664
1547 + },
1548 + {
1549 + "tau": 3.0,
1550 + "escalated_frac": 0.4404296875,
1551 + "residual_disagree": 0.038248697916666664
1552 + },
1553 + {
1554 + "tau": 3.0,
1555 + "escalated_frac": 0.4404296875,
1556 + "residual_disagree": 0.038248697916666664
1557 + },
1558 + {
1559 + "tau": 3.0,
1560 + "escalated_frac": 0.4404296875,
1561 + "residual_disagree": 0.038248697916666664
1562 + },
1563 + {
1564 + "tau": 3.125,
1565 + "escalated_frac": 0.4591471354166667,
1566 + "residual_disagree": 0.034016927083333336
1567 + },
1568 + {
1569 + "tau": 3.125,
1570 + "escalated_frac": 0.4591471354166667,
1571 + "residual_disagree": 0.034016927083333336
1572 + },
1573 + {
1574 + "tau": 3.25,
1575 + "escalated_frac": 0.4685872395833333,
1576 + "residual_disagree": 0.031575520833333336
1577 + },
1578 + {
1579 + "tau": 3.25,
1580 + "escalated_frac": 0.4685872395833333,
1581 + "residual_disagree": 0.031575520833333336
1582 + },
1583 + {
1584 + "tau": 3.25,
1585 + "escalated_frac": 0.4685872395833333,
1586 + "residual_disagree": 0.031575520833333336
1587 + },
1588 + {
1589 + "tau": 3.375,
1590 + "escalated_frac": 0.4855143229166667,
1591 + "residual_disagree": 0.026529947916666668
1592 + },
1593 + {
1594 + "tau": 3.375,
1595 + "escalated_frac": 0.4855143229166667,
1596 + "residual_disagree": 0.026529947916666668
1597 + },
1598 + {
1599 + "tau": 3.5,
1600 + "escalated_frac": 0.4947916666666667,
1601 + "residual_disagree": 0.024251302083333332
1602 + },
1603 + {
1604 + "tau": 3.5,
1605 + "escalated_frac": 0.4947916666666667,
1606 + "residual_disagree": 0.024251302083333332
1607 + },
1608 + {
1609 + "tau": 3.5,
1610 + "escalated_frac": 0.4947916666666667,
1611 + "residual_disagree": 0.024251302083333332
1612 + },
1613 + {
1614 + "tau": 3.625,
1615 + "escalated_frac": 0.5110677083333334,
1616 + "residual_disagree": 0.021321614583333332
1617 + },
1618 + {
1619 + "tau": 3.75,
1620 + "escalated_frac": 0.51708984375,
1621 + "residual_disagree": 0.02001953125
1622 + },
1623 + {
1624 + "tau": 3.75,
1625 + "escalated_frac": 0.51708984375,
1626 + "residual_disagree": 0.02001953125
1627 + },
1628 + {
1629 + "tau": 3.75,
1630 + "escalated_frac": 0.51708984375,
1631 + "residual_disagree": 0.02001953125
1632 + },
1633 + {
1634 + "tau": 3.875,
1635 + "escalated_frac": 0.53173828125,
1636 + "residual_disagree": 0.017415364583333332
1637 + },
1638 + {
1639 + "tau": 3.875,
1640 + "escalated_frac": 0.53173828125,
1641 + "residual_disagree": 0.017415364583333332
1642 + },
1643 + {
1644 + "tau": 4.0,
1645 + "escalated_frac": 0.5384114583333334,
1646 + "residual_disagree": 0.016927083333333332
1647 + },
1648 + {
1649 + "tau": 4.0,
1650 + "escalated_frac": 0.5384114583333334,
1651 + "residual_disagree": 0.016927083333333332
1652 + },
1653 + {
1654 + "tau": 4.125,
1655 + "escalated_frac": 0.5511067708333334,
1656 + "residual_disagree": 0.01416015625
1657 + },
1658 + {
1659 + "tau": 4.25,
1660 + "escalated_frac": 0.5568033854166666,
1661 + "residual_disagree": 0.013997395833333334
1662 + },
1663 + {
1664 + "tau": 4.25,
1665 + "escalated_frac": 0.5568033854166666,
1666 + "residual_disagree": 0.013997395833333334
1667 + },
1668 + {
1669 + "tau": 4.25,
1670 + "escalated_frac": 0.5568033854166666,
1671 + "residual_disagree": 0.013997395833333334
1672 + },
1673 + {
1674 + "tau": 4.375,
1675 + "escalated_frac": 0.5690104166666666,
1676 + "residual_disagree": 0.01318359375
1677 + },
1678 + {
1679 + "tau": 4.5,
1680 + "escalated_frac": 0.5745442708333334,
1681 + "residual_disagree": 0.012858072916666666
1682 + },
1683 + {
1684 + "tau": 4.5,
1685 + "escalated_frac": 0.5745442708333334,
1686 + "residual_disagree": 0.012858072916666666
1687 + },
1688 + {
1689 + "tau": 4.625,
1690 + "escalated_frac": 0.5865885416666666,
1691 + "residual_disagree": 0.011393229166666666
1692 + },
1693 + {
1694 + "tau": 4.75,
1695 + "escalated_frac": 0.5926106770833334,
1696 + "residual_disagree": 0.010904947916666666
1697 + },
1698 + {
1699 + "tau": 4.75,
1700 + "escalated_frac": 0.5926106770833334,
1701 + "residual_disagree": 0.010904947916666666
1702 + },
1703 + {
1704 + "tau": 4.875,
1705 + "escalated_frac": 0.6017252604166666,
1706 + "residual_disagree": 0.009928385416666666
1707 + },
1708 + {
1709 + "tau": 5.0,
1710 + "escalated_frac": 0.6067708333333334,
1711 + "residual_disagree": 0.00927734375
1712 + },
1713 + {
1714 + "tau": 5.0,
1715 + "escalated_frac": 0.6067708333333334,
1716 + "residual_disagree": 0.00927734375
1717 + },
1718 + {
1719 + "tau": 5.125,
1720 + "escalated_frac": 0.61767578125,
1721 + "residual_disagree": 0.0087890625
1722 + },
1723 + {
1724 + "tau": 5.25,
1725 + "escalated_frac": 0.6214192708333334,
1726 + "residual_disagree": 0.008626302083333334
1727 + },
1728 + {
1729 + "tau": 5.25,
1730 + "escalated_frac": 0.6214192708333334,
1731 + "residual_disagree": 0.008626302083333334
1732 + },
1733 + {
1734 + "tau": 5.375,
1735 + "escalated_frac": 0.63134765625,
1736 + "residual_disagree": 0.00732421875
1737 + },
1738 + {
1739 + "tau": 5.5,
1740 + "escalated_frac": 0.6359049479166666,
1741 + "residual_disagree": 0.0068359375
1742 + },
1743 + {
1744 + "tau": 5.5,
1745 + "escalated_frac": 0.6359049479166666,
1746 + "residual_disagree": 0.0068359375
1747 + },
1748 + {
1749 + "tau": 5.75,
1750 + "escalated_frac": 0.6466471354166666,
1751 + "residual_disagree": 0.006510416666666667
1752 + },
1753 + {
1754 + "tau": 5.75,
1755 + "escalated_frac": 0.6466471354166666,
1756 + "residual_disagree": 0.006510416666666667
1757 + },
1758 + {
1759 + "tau": 6.0,
1760 + "escalated_frac": 0.6573893229166666,
1761 + "residual_disagree": 0.00634765625
1762 + },
1763 + {
1764 + "tau": 6.0,
1765 + "escalated_frac": 0.6573893229166666,
1766 + "residual_disagree": 0.00634765625
1767 + },
1768 + {
1769 + "tau": 6.25,
1770 + "escalated_frac": 0.6661783854166666,
1771 + "residual_disagree": 0.005208333333333333
1772 + },
1773 + {
1774 + "tau": 6.375,
1775 + "escalated_frac": 0.67236328125,
1776 + "residual_disagree": 0.004720052083333333
1777 + },
1778 + {
1779 + "tau": 6.5,
1780 + "escalated_frac": 0.6748046875,
1781 + "residual_disagree": 0.004720052083333333
1782 + },
1783 + {
1784 + "tau": 6.625,
1785 + "escalated_frac": 0.6818033854166666,
1786 + "residual_disagree": 0.004069010416666667
1787 + },
1788 + {
1789 + "tau": 6.75,
1790 + "escalated_frac": 0.6863606770833334,
1791 + "residual_disagree": 0.00390625
1792 + },
1793 + {
1794 + "tau": 6.871231155778901,
1795 + "escalated_frac": 0.693359375,
1796 + "residual_disagree": 0.0035807291666666665
1797 + },
1798 + {
1799 + "tau": 7.0,
1800 + "escalated_frac": 0.6964518229166666,
1801 + "residual_disagree": 0.0035807291666666665
1802 + },
1803 + {
1804 + "tau": 7.125,
1805 + "escalated_frac": 0.7024739583333334,
1806 + "residual_disagree": 0.0032552083333333335
1807 + },
1808 + {
1809 + "tau": 7.25,
1810 + "escalated_frac": 0.7052408854166666,
1811 + "residual_disagree": 0.0030924479166666665
1812 + },
1813 + {
1814 + "tau": 7.5,
1815 + "escalated_frac": 0.71337890625,
1816 + "residual_disagree": 0.0027669270833333335
1817 + },
1818 + {
1819 + "tau": 7.5,
1820 + "escalated_frac": 0.71337890625,
1821 + "residual_disagree": 0.0027669270833333335
1822 + },
1823 + {
1824 + "tau": 7.75,
1825 + "escalated_frac": 0.7233072916666666,
1826 + "residual_disagree": 0.00244140625
1827 + },
1828 + {
1829 + "tau": 7.75,
1830 + "escalated_frac": 0.7233072916666666,
1831 + "residual_disagree": 0.00244140625
1832 + },
1833 + {
1834 + "tau": 7.875,
1835 + "escalated_frac": 0.7314453125,
1836 + "residual_disagree": 0.0022786458333333335
1837 + },
1838 + {
1839 + "tau": 8.125,
1840 + "escalated_frac": 0.7373046875,
1841 + "residual_disagree": 0.0022786458333333335
1842 + },
1843 + {
1844 + "tau": 8.25,
1845 + "escalated_frac": 0.73974609375,
1846 + "residual_disagree": 0.0022786458333333335
1847 + },
1848 + {
1849 + "tau": 8.5,
1850 + "escalated_frac": 0.7473958333333334,
1851 + "residual_disagree": 0.0021158854166666665
1852 + },
1853 + {
1854 + "tau": 8.75,
1855 + "escalated_frac": 0.7535807291666666,
1856 + "residual_disagree": 0.0017903645833333333
1857 + },
1858 + {
1859 + "tau": 8.875,
1860 + "escalated_frac": 0.7576497395833334,
1861 + "residual_disagree": 0.0017903645833333333
1862 + },
1863 + {
1864 + "tau": 9.0,
1865 + "escalated_frac": 0.7600911458333334,
1866 + "residual_disagree": 0.0017903645833333333
1867 + },
1868 + {
1869 + "tau": 9.25,
1870 + "escalated_frac": 0.7669270833333334,
1871 + "residual_disagree": 0.0017903645833333333
1872 + },
1873 + {
1874 + "tau": 9.5,
1875 + "escalated_frac": 0.7736002604166666,
1876 + "residual_disagree": 0.0017903645833333333
1877 + },
1878 + {
1879 + "tau": 9.5,
1880 + "escalated_frac": 0.7736002604166666,
1881 + "residual_disagree": 0.0017903645833333333
1882 + },
1883 + {
1884 + "tau": 9.75,
1885 + "escalated_frac": 0.78076171875,
1886 + "residual_disagree": 0.0016276041666666667
1887 + },
1888 + {
1889 + "tau": 10.0,
1890 + "escalated_frac": 0.7862955729166666,
1891 + "residual_disagree": 0.00146484375
1892 + },
1893 + {
1894 + "tau": 10.25,
1895 + "escalated_frac": 0.7926432291666666,
1896 + "residual_disagree": 0.00146484375
1897 + },
1898 + {
1899 + "tau": 10.375,
1900 + "escalated_frac": 0.7967122395833334,
1901 + "residual_disagree": 0.0013020833333333333
1902 + },
1903 + {
1904 + "tau": 10.625,
1905 + "escalated_frac": 0.8035481770833334,
1906 + "residual_disagree": 0.0013020833333333333
1907 + },
1908 + {
1909 + "tau": 10.75,
1910 + "escalated_frac": 0.8059895833333334,
1911 + "residual_disagree": 0.0013020833333333333
1912 + },
1913 + {
1914 + "tau": 10.979271356783897,
1915 + "escalated_frac": 0.81396484375,
1916 + "residual_disagree": 0.0009765625
1917 + },
1918 + {
1919 + "tau": 11.125,
1920 + "escalated_frac": 0.81787109375,
1921 + "residual_disagree": 0.0009765625
1922 + },
1923 + {
1924 + "tau": 11.25,
1925 + "escalated_frac": 0.8203125,
1926 + "residual_disagree": 0.0009765625
1927 + },
1928 + {
1929 + "tau": 11.5,
1930 + "escalated_frac": 0.8269856770833334,
1931 + "residual_disagree": 0.0009765625
1932 + },
1933 + {
1934 + "tau": 11.75,
1935 + "escalated_frac": 0.8336588541666666,
1936 + "residual_disagree": 0.0009765625
1937 + },
1938 + {
1939 + "tau": 11.875,
1940 + "escalated_frac": 0.8369140625,
1941 + "residual_disagree": 0.0009765625
1942 + },
1943 + {
1944 + "tau": 12.125,
1945 + "escalated_frac": 0.8430989583333334,
1946 + "residual_disagree": 0.0008138020833333334
1947 + },
1948 + {
1949 + "tau": 12.364949748743697,
1950 + "escalated_frac": 0.84912109375,
1951 + "residual_disagree": 0.0006510416666666666
1952 + },
1953 + {
1954 + "tau": 12.5,
1955 + "escalated_frac": 0.8518880208333334,
1956 + "residual_disagree": 0.0006510416666666666
1957 + },
1958 + {
1959 + "tau": 12.75,
1960 + "escalated_frac": 0.8583984375,
1961 + "residual_disagree": 0.00048828125
1962 + },
1963 + {
1964 + "tau": 13.0,
1965 + "escalated_frac": 0.8640950520833334,
1966 + "residual_disagree": 0.0003255208333333333
1967 + },
1968 + {
1969 + "tau": 13.25,
1970 + "escalated_frac": 0.8684895833333334,
1971 + "residual_disagree": 0.0003255208333333333
1972 + },
1973 + {
1974 + "tau": 13.5,
1975 + "escalated_frac": 0.87353515625,
1976 + "residual_disagree": 0.00016276041666666666
1977 + },
1978 + {
1979 + "tau": 13.75,
1980 + "escalated_frac": 0.8790690104166666,
1981 + "residual_disagree": 0.0
1982 + },
1983 + {
1984 + "tau": 13.875,
1985 + "escalated_frac": 0.8834635416666666,
1986 + "residual_disagree": 0.0
1987 + },
1988 + {
1989 + "tau": 14.125,
1990 + "escalated_frac": 0.8885091145833334,
1991 + "residual_disagree": 0.0
1992 + },
1993 + {
1994 + "tau": 14.25,
1995 + "escalated_frac": 0.8917643229166666,
1996 + "residual_disagree": 0.0
1997 + },
1998 + {
1999 + "tau": 14.5,
2000 + "escalated_frac": 0.8963216145833334,
2001 + "residual_disagree": 0.0
2002 + },
2003 + {
2004 + "tau": 14.75,
2005 + "escalated_frac": 0.9021809895833334,
2006 + "residual_disagree": 0.0
2007 + },
2008 + {
2009 + "tau": 15.0,
2010 + "escalated_frac": 0.9075520833333334,
2011 + "residual_disagree": 0.0
2012 + },
2013 + {
2014 + "tau": 15.27763819095469,
2015 + "escalated_frac": 0.91455078125,
2016 + "residual_disagree": 0.0
2017 + },
2018 + {
2019 + "tau": 15.625,
2020 + "escalated_frac": 0.9189453125,
2021 + "residual_disagree": 0.0
2022 + },
2023 + {
2024 + "tau": 16.0,
2025 + "escalated_frac": 0.9236653645833334,
2026 + "residual_disagree": 0.0
2027 + },
2028 + {
2029 + "tau": 16.25,
2030 + "escalated_frac": 0.92822265625,
2031 + "residual_disagree": 0.0
2032 + },
2033 + {
2034 + "tau": 16.75,
2035 + "escalated_frac": 0.9339192708333334,
2036 + "residual_disagree": 0.0
2037 + },
2038 + {
2039 + "tau": 17.25,
2040 + "escalated_frac": 0.9392903645833334,
2041 + "residual_disagree": 0.0
2042 + },
2043 + {
2044 + "tau": 17.625,
2045 + "escalated_frac": 0.9440104166666666,
2046 + "residual_disagree": 0.0
2047 + },
2048 + {
2049 + "tau": 17.875,
2050 + "escalated_frac": 0.9484049479166666,
2051 + "residual_disagree": 0.0
2052 + },
2053 + {
2054 + "tau": 18.375,
2055 + "escalated_frac": 0.9542643229166666,
2056 + "residual_disagree": 0.0
2057 + },
2058 + {
2059 + "tau": 18.875,
2060 + "escalated_frac": 0.95947265625,
2061 + "residual_disagree": 0.0
2062 + },
2063 + {
2064 + "tau": 19.375,
2065 + "escalated_frac": 0.9640299479166666,
2066 + "residual_disagree": 0.0
2067 + },
2068 + {
2069 + "tau": 20.0,
2070 + "escalated_frac": 0.9695638020833334,
2071 + "residual_disagree": 0.0
2072 + },
2073 + {
2074 + "tau": 20.375,
2075 + "escalated_frac": 0.97412109375,
2076 + "residual_disagree": 0.0
2077 + },
2078 + {
2079 + "tau": 21.125,
2080 + "escalated_frac": 0.9793294270833334,
2081 + "residual_disagree": 0.0
2082 + },
2083 + {
2084 + "tau": 22.173994974874404,
2085 + "escalated_frac": 0.98486328125,
2086 + "residual_disagree": 0.0
2087 + },
2088 + {
2089 + "tau": 23.53266331658301,
2090 + "escalated_frac": 0.9899088541666666,
2091 + "residual_disagree": 0.0
2092 + },
2093 + {
2094 + "tau": 25.391331658291392,
2095 + "escalated_frac": 0.9949544270833334,
2096 + "residual_disagree": 0.0
2097 + },
2098 + {
2099 + "tau": 32.375,
2100 + "escalated_frac": 0.9998372395833334,
2101 + "residual_disagree": 0.0
2102 + }
2103 + ],
2104 + "per_domain": {
2105 + "code": {
2106 + "agreement_rate": 0.728515625,
2107 + "auroc": 0.8375484598916041
2108 + },
2109 + "math": {
2110 + "agreement_rate": 0.8193359375,
2111 + "auroc": 0.8957188416068035
2112 + },
2113 + "chat": {
2114 + "agreement_rate": 0.6669921875,
2115 + "auroc": 0.8009257072686913
2116 + },
2117 + "french": {
2118 + "agreement_rate": 0.728515625,
2119 + "auroc": 0.8465605531660463
2120 + },
2121 + "english": {
2122 + "agreement_rate": 0.6875,
2123 + "auroc": 0.8543834339488636
2124 + },
2125 + "reasoning": {
2126 + "agreement_rate": 0.7548828125,
2127 + "auroc": 0.8626838055282106
2128 + }
2129 + },
2130 + "escalation_frac_for_99pct": 0.6017252604166666
2131 + },
2132 + "4": {
2133 + "bits": 4,
2134 + "n_positions": 6144,
2135 + "agreement_rate": 0.8740234375,
2136 + "mean_kl_ref_q": 0.19848966598510742,
2137 + "auroc_margin_predicts_disagreement": 0.8982447466304814,
2138 + "median_margin_agree": 6.5,
2139 + "median_margin_disagree": 0.75,
2140 + "escalation_curve": [
2141 + {
2142 + "tau": 0.0,
2143 + "escalated_frac": 0.0,
2144 + "residual_disagree": 0.1259765625
2145 + },
2146 + {
2147 + "tau": 0.0,
2148 + "escalated_frac": 0.0,
2149 + "residual_disagree": 0.1259765625
2150 + },
2151 + {
2152 + "tau": 0.0,
2153 + "escalated_frac": 0.0,
2154 + "residual_disagree": 0.1259765625
2155 + },
2156 + {
2157 + "tau": 0.125,
2158 + "escalated_frac": 0.014485677083333334,
2159 + "residual_disagree": 0.11783854166666667
2160 + },
2161 + {
2162 + "tau": 0.125,
2163 + "escalated_frac": 0.014485677083333334,
2164 + "residual_disagree": 0.11783854166666667
2165 + },
2166 + {
2167 + "tau": 0.125,
2168 + "escalated_frac": 0.014485677083333334,
2169 + "residual_disagree": 0.11783854166666667
2170 + },
2171 + {
2172 + "tau": 0.25,
2173 + "escalated_frac": 0.0263671875,
2174 + "residual_disagree": 0.1083984375
2175 + },
2176 + {
2177 + "tau": 0.25,
2178 + "escalated_frac": 0.0263671875,
2179 + "residual_disagree": 0.1083984375
2180 + },
2181 + {
2182 + "tau": 0.25,
2183 + "escalated_frac": 0.0263671875,
2184 + "residual_disagree": 0.1083984375
2185 + },
2186 + {
2187 + "tau": 0.25,
2188 + "escalated_frac": 0.0263671875,
2189 + "residual_disagree": 0.1083984375
2190 + },
2191 + {
2192 + "tau": 0.25,
2193 + "escalated_frac": 0.0263671875,
2194 + "residual_disagree": 0.1083984375
2195 + },
2196 + {
2197 + "tau": 0.375,
2198 + "escalated_frac": 0.053059895833333336,
2199 + "residual_disagree": 0.09309895833333333
2200 + },
2201 + {
2202 + "tau": 0.375,
2203 + "escalated_frac": 0.053059895833333336,
2204 + "residual_disagree": 0.09309895833333333
2205 + },
2206 + {
2207 + "tau": 0.5,
2208 + "escalated_frac": 0.06266276041666667,
2209 + "residual_disagree": 0.08756510416666667
2210 + },
2211 + {
2212 + "tau": 0.5,
2213 + "escalated_frac": 0.06266276041666667,
2214 + "residual_disagree": 0.08756510416666667
2215 + },
2216 + {
2217 + "tau": 0.5,
2218 + "escalated_frac": 0.06266276041666667,
2219 + "residual_disagree": 0.08756510416666667
2220 + },
2221 + {
2222 + "tau": 0.5,
2223 + "escalated_frac": 0.06266276041666667,
2224 + "residual_disagree": 0.08756510416666667
2225 + },
2226 + {
2227 + "tau": 0.5,
2228 + "escalated_frac": 0.06266276041666667,
2229 + "residual_disagree": 0.08756510416666667
2230 + },
2231 + {
2232 + "tau": 0.625,
2233 + "escalated_frac": 0.08821614583333333,
2234 + "residual_disagree": 0.07682291666666667
2235 + },
2236 + {
2237 + "tau": 0.625,
2238 + "escalated_frac": 0.08821614583333333,
2239 + "residual_disagree": 0.07682291666666667
2240 + },
2241 + {
2242 + "tau": 0.75,
2243 + "escalated_frac": 0.09781901041666667,
2244 + "residual_disagree": 0.07291666666666667
2245 + },
2246 + {
2247 + "tau": 0.75,
2248 + "escalated_frac": 0.09781901041666667,
2249 + "residual_disagree": 0.07291666666666667
2250 + },
2251 + {
2252 + "tau": 0.75,
2253 + "escalated_frac": 0.09781901041666667,
2254 + "residual_disagree": 0.07291666666666667
2255 + },
2256 + {
2257 + "tau": 0.75,
2258 + "escalated_frac": 0.09781901041666667,
2259 + "residual_disagree": 0.07291666666666667
2260 + },
2261 + {
2262 + "tau": 0.75,
2263 + "escalated_frac": 0.09781901041666667,
2264 + "residual_disagree": 0.07291666666666667
2265 + },
2266 + {
2267 + "tau": 0.875,
2268 + "escalated_frac": 0.12337239583333333,
2269 + "residual_disagree": 0.062174479166666664
2270 + },
2271 + {
2272 + "tau": 0.875,
2273 + "escalated_frac": 0.12337239583333333,
2274 + "residual_disagree": 0.062174479166666664
2275 + },
2276 + {
2277 + "tau": 1.0,
2278 + "escalated_frac": 0.1318359375,
2279 + "residual_disagree": 0.05859375
2280 + },
2281 + {
2282 + "tau": 1.0,
2283 + "escalated_frac": 0.1318359375,
2284 + "residual_disagree": 0.05859375
2285 + },
2286 + {
2287 + "tau": 1.0,
2288 + "escalated_frac": 0.1318359375,
2289 + "residual_disagree": 0.05859375
2290 + },
2291 + {
2292 + "tau": 1.0,
2293 + "escalated_frac": 0.1318359375,
2294 + "residual_disagree": 0.05859375
2295 + },
2296 + {
2297 + "tau": 1.125,
2298 + "escalated_frac": 0.15543619791666666,
2299 + "residual_disagree": 0.050455729166666664
2300 + },
2301 + {
2302 + "tau": 1.125,
2303 + "escalated_frac": 0.15543619791666666,
2304 + "residual_disagree": 0.050455729166666664
2305 + },
2306 + {
2307 + "tau": 1.25,
2308 + "escalated_frac": 0.16292317708333334,
2309 + "residual_disagree": 0.047526041666666664
2310 + },
2311 + {
2312 + "tau": 1.25,
2313 + "escalated_frac": 0.16292317708333334,
2314 + "residual_disagree": 0.047526041666666664
2315 + },
2316 + {
2317 + "tau": 1.25,
2318 + "escalated_frac": 0.16292317708333334,
2319 + "residual_disagree": 0.047526041666666664
2320 + },
2321 + {
2322 + "tau": 1.25,
2323 + "escalated_frac": 0.16292317708333334,
2324 + "residual_disagree": 0.047526041666666664
2325 + },
2326 + {
2327 + "tau": 1.25,
2328 + "escalated_frac": 0.16292317708333334,
2329 + "residual_disagree": 0.047526041666666664
2330 + },
2331 + {
2332 + "tau": 1.375,
2333 + "escalated_frac": 0.18880208333333334,
2334 + "residual_disagree": 0.040364583333333336
2335 + },
2336 + {
2337 + "tau": 1.375,
2338 + "escalated_frac": 0.18880208333333334,
2339 + "residual_disagree": 0.040364583333333336
2340 + },
2341 + {
2342 + "tau": 1.5,
2343 + "escalated_frac": 0.19612630208333334,
2344 + "residual_disagree": 0.0390625
2345 + },
2346 + {
2347 + "tau": 1.5,
2348 + "escalated_frac": 0.19612630208333334,
2349 + "residual_disagree": 0.0390625
2350 + },
2351 + {
2352 + "tau": 1.5,
2353 + "escalated_frac": 0.19612630208333334,
2354 + "residual_disagree": 0.0390625
2355 + },
2356 + {
2357 + "tau": 1.5,
2358 + "escalated_frac": 0.19612630208333334,
2359 + "residual_disagree": 0.0390625
2360 + },
2361 + {
2362 + "tau": 1.625,
2363 + "escalated_frac": 0.21663411458333334,
2364 + "residual_disagree": 0.03271484375
2365 + },
2366 + {
2367 + "tau": 1.75,
2368 + "escalated_frac": 0.22412109375,
2369 + "residual_disagree": 0.030436197916666668
2370 + },
2371 + {
2372 + "tau": 1.75,
2373 + "escalated_frac": 0.22412109375,
2374 + "residual_disagree": 0.030436197916666668
2375 + },
2376 + {
2377 + "tau": 1.75,
2378 + "escalated_frac": 0.22412109375,
2379 + "residual_disagree": 0.030436197916666668
2380 + },
2381 + {
2382 + "tau": 1.75,
2383 + "escalated_frac": 0.22412109375,
2384 + "residual_disagree": 0.030436197916666668
2385 + },
2386 + {
2387 + "tau": 1.875,
2388 + "escalated_frac": 0.24235026041666666,
2389 + "residual_disagree": 0.027180989583333332
2390 + },
2391 + {
2392 + "tau": 2.0,
2393 + "escalated_frac": 0.24886067708333334,
2394 + "residual_disagree": 0.026041666666666668
2395 + },
2396 + {
2397 + "tau": 2.0,
2398 + "escalated_frac": 0.24886067708333334,
2399 + "residual_disagree": 0.026041666666666668
2400 + },
2401 + {
2402 + "tau": 2.0,
2403 + "escalated_frac": 0.24886067708333334,
2404 + "residual_disagree": 0.026041666666666668
2405 + },
2406 + {
2407 + "tau": 2.0,
2408 + "escalated_frac": 0.24886067708333334,
2409 + "residual_disagree": 0.026041666666666668
2410 + },
2411 + {
2412 + "tau": 2.125,
2413 + "escalated_frac": 0.2688802083333333,
2414 + "residual_disagree": 0.02197265625
2415 + },
2416 + {
2417 + "tau": 2.125,
2418 + "escalated_frac": 0.2688802083333333,
2419 + "residual_disagree": 0.02197265625
2420 + },
2421 + {
2422 + "tau": 2.25,
2423 + "escalated_frac": 0.2766927083333333,
2424 + "residual_disagree": 0.020182291666666668
2425 + },
2426 + {
2427 + "tau": 2.25,
2428 + "escalated_frac": 0.2766927083333333,
2429 + "residual_disagree": 0.020182291666666668
2430 + },
2431 + {
2432 + "tau": 2.25,
2433 + "escalated_frac": 0.2766927083333333,
2434 + "residual_disagree": 0.020182291666666668
2435 + },
2436 + {
2437 + "tau": 2.375,
2438 + "escalated_frac": 0.2939453125,
2439 + "residual_disagree": 0.017903645833333332
2440 + },
2441 + {
2442 + "tau": 2.5,
2443 + "escalated_frac": 0.3004557291666667,
2444 + "residual_disagree": 0.017252604166666668
2445 + },
2446 + {
2447 + "tau": 2.5,
2448 + "escalated_frac": 0.3004557291666667,
2449 + "residual_disagree": 0.017252604166666668
2450 + },
2451 + {
2452 + "tau": 2.5,
2453 + "escalated_frac": 0.3004557291666667,
2454 + "residual_disagree": 0.017252604166666668
2455 + },
2456 + {
2457 + "tau": 2.5,
2458 + "escalated_frac": 0.3004557291666667,
2459 + "residual_disagree": 0.017252604166666668
2460 + },
2461 + {
2462 + "tau": 2.625,
2463 + "escalated_frac": 0.3196614583333333,
2464 + "residual_disagree": 0.014811197916666666
2465 + },
2466 + {
2467 + "tau": 2.75,
2468 + "escalated_frac": 0.3263346354166667,
2469 + "residual_disagree": 0.013834635416666666
2470 + },
2471 + {
2472 + "tau": 2.75,
2473 + "escalated_frac": 0.3263346354166667,
2474 + "residual_disagree": 0.013834635416666666
2475 + },
2476 + {
2477 + "tau": 2.75,
2478 + "escalated_frac": 0.3263346354166667,
2479 + "residual_disagree": 0.013834635416666666
2480 + },
2481 + {
2482 + "tau": 2.75,
2483 + "escalated_frac": 0.3263346354166667,
2484 + "residual_disagree": 0.013834635416666666
2485 + },
2486 + {
2487 + "tau": 2.875,
2488 + "escalated_frac": 0.3448893229166667,
2489 + "residual_disagree": 0.01123046875
2490 + },
2491 + {
2492 + "tau": 3.0,
2493 + "escalated_frac": 0.3512369791666667,
2494 + "residual_disagree": 0.01025390625
2495 + },
2496 + {
2497 + "tau": 3.0,
2498 + "escalated_frac": 0.3512369791666667,
2499 + "residual_disagree": 0.01025390625
2500 + },
2501 + {
2502 + "tau": 3.0,
2503 + "escalated_frac": 0.3512369791666667,
2504 + "residual_disagree": 0.01025390625
2505 + },
2506 + {
2507 + "tau": 3.125,
2508 + "escalated_frac": 0.3658854166666667,
2509 + "residual_disagree": 0.009114583333333334
2510 + },
2511 + {
2512 + "tau": 3.25,
2513 + "escalated_frac": 0.37060546875,
2514 + "residual_disagree": 0.00830078125
2515 + },
2516 + {
2517 + "tau": 3.25,
2518 + "escalated_frac": 0.37060546875,
2519 + "residual_disagree": 0.00830078125
2520 + },
2521 + {
2522 + "tau": 3.25,
2523 + "escalated_frac": 0.37060546875,
2524 + "residual_disagree": 0.00830078125
2525 + },
2526 + {
2527 + "tau": 3.375,
2528 + "escalated_frac": 0.38623046875,
2529 + "residual_disagree": 0.007161458333333333
2530 + },
2531 + {
2532 + "tau": 3.5,
2533 + "escalated_frac": 0.3914388020833333,
2534 + "residual_disagree": 0.006673177083333333
2535 + },
2536 + {
2537 + "tau": 3.5,
2538 + "escalated_frac": 0.3914388020833333,
2539 + "residual_disagree": 0.006673177083333333
2540 + },
2541 + {
2542 + "tau": 3.5,
2543 + "escalated_frac": 0.3914388020833333,
2544 + "residual_disagree": 0.006673177083333333
2545 + },
2546 + {
2547 + "tau": 3.5521356783919487,
2548 + "escalated_frac": 0.4070638020833333,
2549 + "residual_disagree": 0.00537109375
2550 + },
2551 + {
2552 + "tau": 3.75,
2553 + "escalated_frac": 0.4112955729166667,
2554 + "residual_disagree": 0.005045572916666667
2555 + },
2556 + {
2557 + "tau": 3.75,
2558 + "escalated_frac": 0.4112955729166667,
2559 + "residual_disagree": 0.005045572916666667
2560 + },
2561 + {
2562 + "tau": 3.75,
2563 + "escalated_frac": 0.4112955729166667,
2564 + "residual_disagree": 0.005045572916666667
2565 + },
2566 + {
2567 + "tau": 3.875,
2568 + "escalated_frac": 0.4231770833333333,
2569 + "residual_disagree": 0.004231770833333333
2570 + },
2571 + {
2572 + "tau": 4.0,
2573 + "escalated_frac": 0.42724609375,
2574 + "residual_disagree": 0.004231770833333333
2575 + },
2576 + {
2577 + "tau": 4.0,
2578 + "escalated_frac": 0.42724609375,
2579 + "residual_disagree": 0.004231770833333333
2580 + },
2581 + {
2582 + "tau": 4.125,
2583 + "escalated_frac": 0.4386393229166667,
2584 + "residual_disagree": 0.0037434895833333335
2585 + },
2586 + {
2587 + "tau": 4.25,
2588 + "escalated_frac": 0.44287109375,
2589 + "residual_disagree": 0.0037434895833333335
2590 + },
2591 + {
2592 + "tau": 4.25,
2593 + "escalated_frac": 0.44287109375,
2594 + "residual_disagree": 0.0037434895833333335
2595 + },
2596 + {
2597 + "tau": 4.375,
2598 + "escalated_frac": 0.4558919270833333,
2599 + "residual_disagree": 0.0029296875
2600 + },
2601 + {
2602 + "tau": 4.5,
2603 + "escalated_frac": 0.4591471354166667,
2604 + "residual_disagree": 0.0029296875
2605 + },
2606 + {
2607 + "tau": 4.5,
2608 + "escalated_frac": 0.4591471354166667,
2609 + "residual_disagree": 0.0029296875
2610 + },
2611 + {
2612 + "tau": 4.625,
2613 + "escalated_frac": 0.4698893229166667,
2614 + "residual_disagree": 0.00244140625
2615 + },
2616 + {
2617 + "tau": 4.75,
2618 + "escalated_frac": 0.4737955729166667,
2619 + "residual_disagree": 0.00244140625
2620 + },
2621 + {
2622 + "tau": 4.75,
2623 + "escalated_frac": 0.4737955729166667,
2624 + "residual_disagree": 0.00244140625
2625 + },
2626 + {
2627 + "tau": 4.875,
2628 + "escalated_frac": 0.4850260416666667,
2629 + "residual_disagree": 0.001953125
2630 + },
2631 + {
2632 + "tau": 5.0,
2633 + "escalated_frac": 0.4889322916666667,
2634 + "residual_disagree": 0.001953125
2635 + },
2636 + {
2637 + "tau": 5.0,
2638 + "escalated_frac": 0.4889322916666667,
2639 + "residual_disagree": 0.001953125
2640 + },
2641 + {
2642 + "tau": 5.125,
2643 + "escalated_frac": 0.49853515625,
2644 + "residual_disagree": 0.001953125
2645 + },
2646 + {
2647 + "tau": 5.25,
2648 + "escalated_frac": 0.50390625,
2649 + "residual_disagree": 0.001953125
2650 + },
2651 + {
2652 + "tau": 5.25,
2653 + "escalated_frac": 0.50390625,
2654 + "residual_disagree": 0.001953125
2655 + },
2656 + {
2657 + "tau": 5.5,
2658 + "escalated_frac": 0.5149739583333334,
2659 + "residual_disagree": 0.0017903645833333333
2660 + },
2661 + {
2662 + "tau": 5.5,
2663 + "escalated_frac": 0.5149739583333334,
2664 + "residual_disagree": 0.0017903645833333333
2665 + },
2666 + {
2667 + "tau": 5.625,
2668 + "escalated_frac": 0.5255533854166666,
2669 + "residual_disagree": 0.0016276041666666667
2670 + },
2671 + {
2672 + "tau": 5.75,
2673 + "escalated_frac": 0.5284830729166666,
2674 + "residual_disagree": 0.0016276041666666667
2675 + },
2676 + {
2677 + "tau": 5.875,
2678 + "escalated_frac": 0.537109375,
2679 + "residual_disagree": 0.0013020833333333333
2680 + },
2681 + {
2682 + "tau": 6.0,
2683 + "escalated_frac": 0.5413411458333334,
2684 + "residual_disagree": 0.0013020833333333333
2685 + },
2686 + {
2687 + "tau": 6.125,
2688 + "escalated_frac": 0.5472005208333334,
2689 + "residual_disagree": 0.0011393229166666667
2690 + },
2691 + {
2692 + "tau": 6.25,
2693 + "escalated_frac": 0.5491536458333334,
2694 + "residual_disagree": 0.0011393229166666667
2695 + },
2696 + {
2697 + "tau": 6.375,
2698 + "escalated_frac": 0.55712890625,
2699 + "residual_disagree": 0.0011393229166666667
2700 + },
2701 + {
2702 + "tau": 6.5,
2703 + "escalated_frac": 0.5597330729166666,
2704 + "residual_disagree": 0.0011393229166666667
2705 + },
2706 + {
2707 + "tau": 6.625,
2708 + "escalated_frac": 0.5665690104166666,
2709 + "residual_disagree": 0.0009765625
2710 + },
2711 + {
2712 + "tau": 6.75,
2713 + "escalated_frac": 0.57080078125,
2714 + "residual_disagree": 0.0009765625
2715 + },
2716 + {
2717 + "tau": 6.75,
2718 + "escalated_frac": 0.57080078125,
2719 + "residual_disagree": 0.0009765625
2720 + },
2721 + {
2722 + "tau": 7.0,
2723 + "escalated_frac": 0.58154296875,
2724 + "residual_disagree": 0.00048828125
2725 + },
2726 + {
2727 + "tau": 7.0,
2728 + "escalated_frac": 0.58154296875,
2729 + "residual_disagree": 0.00048828125
2730 + },
2731 + {
2732 + "tau": 7.197864321608051,
2733 + "escalated_frac": 0.5929361979166666,
2734 + "residual_disagree": 0.0003255208333333333
2735 + },
2736 + {
2737 + "tau": 7.25,
2738 + "escalated_frac": 0.5929361979166666,
2739 + "residual_disagree": 0.0003255208333333333
2740 + },
2741 + {
2742 + "tau": 7.375,
2743 + "escalated_frac": 0.6013997395833334,
2744 + "residual_disagree": 0.0003255208333333333
2745 + },
2746 + {
2747 + "tau": 7.5,
2748 + "escalated_frac": 0.6046549479166666,
2749 + "residual_disagree": 0.0003255208333333333
2750 + },
2751 + {
2752 + "tau": 7.75,
2753 + "escalated_frac": 0.6123046875,
2754 + "residual_disagree": 0.0003255208333333333
2755 + },
2756 + {
2757 + "tau": 7.875,
2758 + "escalated_frac": 0.6178385416666666,
2759 + "residual_disagree": 0.0003255208333333333
2760 + },
2761 + {
2762 + "tau": 8.0,
2763 + "escalated_frac": 0.6214192708333334,
2764 + "residual_disagree": 0.0003255208333333333
2765 + },
2766 + {
2767 + "tau": 8.125,
2768 + "escalated_frac": 0.6272786458333334,
2769 + "residual_disagree": 0.0003255208333333333
2770 + },
2771 + {
2772 + "tau": 8.25,
2773 + "escalated_frac": 0.6316731770833334,
2774 + "residual_disagree": 0.0003255208333333333
2775 + },
2776 + {
2777 + "tau": 8.375,
2778 + "escalated_frac": 0.6376953125,
2779 + "residual_disagree": 0.0003255208333333333
2780 + },
2781 + {
2782 + "tau": 8.5,
2783 + "escalated_frac": 0.640625,
2784 + "residual_disagree": 0.0003255208333333333
2785 + },
2786 + {
2787 + "tau": 8.625,
2788 + "escalated_frac": 0.6466471354166666,
2789 + "residual_disagree": 0.0003255208333333333
2790 + },
2791 + {
2792 + "tau": 8.75,
2793 + "escalated_frac": 0.650390625,
2794 + "residual_disagree": 0.0003255208333333333
2795 + },
2796 + {
2797 + "tau": 8.875,
2798 + "escalated_frac": 0.6555989583333334,
2799 + "residual_disagree": 0.0003255208333333333
2800 + },
2801 + {
2802 + "tau": 9.0,
2803 + "escalated_frac": 0.6598307291666666,
2804 + "residual_disagree": 0.0003255208333333333
2805 + },
2806 + {
2807 + "tau": 9.125,
2808 + "escalated_frac": 0.6653645833333334,
2809 + "residual_disagree": 0.0
2810 + },
2811 + {
2812 + "tau": 9.25,
2813 + "escalated_frac": 0.6686197916666666,
2814 + "residual_disagree": 0.0
2815 + },
2816 + {
2817 + "tau": 9.5,
2818 + "escalated_frac": 0.6770833333333334,
2819 + "residual_disagree": 0.0
2820 + },
2821 + {
2822 + "tau": 9.5,
2823 + "escalated_frac": 0.6770833333333334,
2824 + "residual_disagree": 0.0
2825 + },
2826 + {
2827 + "tau": 9.75,
2828 + "escalated_frac": 0.6861979166666666,
2829 + "residual_disagree": 0.0
2830 + },
2831 + {
2832 + "tau": 9.9962311557789,
2833 + "escalated_frac": 0.693359375,
2834 + "residual_disagree": 0.0
2835 + },
2836 + {
2837 + "tau": 10.125,
2838 + "escalated_frac": 0.6971028645833334,
2839 + "residual_disagree": 0.0
2840 + },
2841 + {
2842 + "tau": 10.25,
2843 + "escalated_frac": 0.6988932291666666,
2844 + "residual_disagree": 0.0
2845 + },
2846 + {
2847 + "tau": 10.5,
2848 + "escalated_frac": 0.7068684895833334,
2849 + "residual_disagree": 0.0
2850 + },
2851 + {
2852 + "tau": 10.625,
2853 + "escalated_frac": 0.7120768229166666,
2854 + "residual_disagree": 0.0
2855 + },
2856 + {
2857 + "tau": 10.75,
2858 + "escalated_frac": 0.71435546875,
2859 + "residual_disagree": 0.0
2860 + },
2861 + {
2862 + "tau": 11.0,
2863 + "escalated_frac": 0.7220052083333334,
2864 + "residual_disagree": 0.0
2865 + },
2866 + {
2867 + "tau": 11.125,
2868 + "escalated_frac": 0.7275390625,
2869 + "residual_disagree": 0.0
2870 + },
2871 + {
2872 + "tau": 11.375,
2873 + "escalated_frac": 0.7327473958333334,
2874 + "residual_disagree": 0.0
2875 + },
2876 + {
2877 + "tau": 11.5,
2878 + "escalated_frac": 0.7360026041666666,
2879 + "residual_disagree": 0.0
2880 + },
2881 + {
2882 + "tau": 11.75,
2883 + "escalated_frac": 0.7410481770833334,
2884 + "residual_disagree": 0.0
2885 + },
2886 + {
2887 + "tau": 12.0,
2888 + "escalated_frac": 0.7478841145833334,
2889 + "residual_disagree": 0.0
2890 + },
2891 + {
2892 + "tau": 12.25,
2893 + "escalated_frac": 0.7529296875,
2894 + "residual_disagree": 0.0
2895 + },
2896 + {
2897 + "tau": 12.375,
2898 + "escalated_frac": 0.7566731770833334,
2899 + "residual_disagree": 0.0
2900 + },
2901 + {
2902 + "tau": 12.75,
2903 + "escalated_frac": 0.7630208333333334,
2904 + "residual_disagree": 0.0
2905 + },
2906 + {
2907 + "tau": 12.875,
2908 + "escalated_frac": 0.7667643229166666,
2909 + "residual_disagree": 0.0
2910 + },
2911 + {
2912 + "tau": 13.125,
2913 + "escalated_frac": 0.7734375,
2914 + "residual_disagree": 0.0
2915 + },
2916 + {
2917 + "tau": 13.343592964824097,
2918 + "escalated_frac": 0.77880859375,
2919 + "residual_disagree": 0.0
2920 + },
2921 + {
2922 + "tau": 13.5,
2923 + "escalated_frac": 0.78173828125,
2924 + "residual_disagree": 0.0
2925 + },
2926 + {
2927 + "tau": 13.625,
2928 + "escalated_frac": 0.7880859375,
2929 + "residual_disagree": 0.0
2930 + },
2931 + {
2932 + "tau": 13.875,
2933 + "escalated_frac": 0.79345703125,
2934 + "residual_disagree": 0.0
2935 + },
2936 + {
2937 + "tau": 14.0,
2938 + "escalated_frac": 0.7972005208333334,
2939 + "residual_disagree": 0.0
2940 + },
2941 + {
2942 + "tau": 14.25,
2943 + "escalated_frac": 0.8025716145833334,
2944 + "residual_disagree": 0.0
2945 + },
2946 + {
2947 + "tau": 14.375,
2948 + "escalated_frac": 0.8079427083333334,
2949 + "residual_disagree": 0.0
2950 + },
2951 + {
2952 + "tau": 14.5,
2953 + "escalated_frac": 0.8103841145833334,
2954 + "residual_disagree": 0.0
2955 + },
2956 + {
2957 + "tau": 14.75,
2958 + "escalated_frac": 0.8170572916666666,
2959 + "residual_disagree": 0.0
2960 + },
2961 + {
2962 + "tau": 14.875,
2963 + "escalated_frac": 0.8212890625,
2964 + "residual_disagree": 0.0
2965 + },
2966 + {
2967 + "tau": 15.180276381909607,
2968 + "escalated_frac": 0.8291015625,
2969 + "residual_disagree": 0.0
2970 + },
2971 + {
2972 + "tau": 15.5,
2973 + "escalated_frac": 0.8331705729166666,
2974 + "residual_disagree": 0.0
2975 + },
2976 + {
2977 + "tau": 15.75,
2978 + "escalated_frac": 0.8377278645833334,
2979 + "residual_disagree": 0.0
2980 + },
2981 + {
2982 + "tau": 16.0,
2983 + "escalated_frac": 0.8435872395833334,
2984 + "residual_disagree": 0.0
2985 + },
2986 + {
2987 + "tau": 16.125,
2988 + "escalated_frac": 0.84814453125,
2989 + "residual_disagree": 0.0
2990 + },
2991 + {
2992 + "tau": 16.375,
2993 + "escalated_frac": 0.8528645833333334,
2994 + "residual_disagree": 0.0
2995 + },
2996 + {
2997 + "tau": 16.625,
2998 + "escalated_frac": 0.8583984375,
2999 + "residual_disagree": 0.0
3000 + },
3001 + {
3002 + "tau": 16.815954773869294,
3003 + "escalated_frac": 0.8642578125,
3004 + "residual_disagree": 0.0
3005 + },
3006 + {
3007 + "tau": 17.125,
3008 + "escalated_frac": 0.8680013020833334,
3009 + "residual_disagree": 0.0
3010 + },
3011 + {
3012 + "tau": 17.375,
3013 + "escalated_frac": 0.8738606770833334,
3014 + "residual_disagree": 0.0
3015 + },
3016 + {
3017 + "tau": 17.625,
3018 + "escalated_frac": 0.8782552083333334,
3019 + "residual_disagree": 0.0
3020 + },
3021 + {
3022 + "tau": 17.75062814070361,
3023 + "escalated_frac": 0.8844401041666666,
3024 + "residual_disagree": 0.0
3025 + },
3026 + {
3027 + "tau": 18.125,
3028 + "escalated_frac": 0.8880208333333334,
3029 + "residual_disagree": 0.0
3030 + },
3031 + {
3032 + "tau": 18.5,
3033 + "escalated_frac": 0.8937174479166666,
3034 + "residual_disagree": 0.0
3035 + },
3036 + {
3037 + "tau": 18.875,
3038 + "escalated_frac": 0.8990885416666666,
3039 + "residual_disagree": 0.0
3040 + },
3041 + {
3042 + "tau": 19.125,
3043 + "escalated_frac": 0.9025065104166666,
3044 + "residual_disagree": 0.0
3045 + },
3046 + {
3047 + "tau": 19.29396984924631,
3048 + "escalated_frac": 0.9095052083333334,
3049 + "residual_disagree": 0.0
3050 + },
3051 + {
3052 + "tau": 19.625,
3053 + "escalated_frac": 0.9138997395833334,
3054 + "residual_disagree": 0.0
3055 + },
3056 + {
3057 + "tau": 19.875,
3058 + "escalated_frac": 0.9181315104166666,
3059 + "residual_disagree": 0.0
3060 + },
3061 + {
3062 + "tau": 20.25,
3063 + "escalated_frac": 0.9235026041666666,
3064 + "residual_disagree": 0.0
3065 + },
3066 + {
3067 + "tau": 20.625,
3068 + "escalated_frac": 0.9285481770833334,
3069 + "residual_disagree": 0.0
3070 + },
3071 + {
3072 + "tau": 21.0,
3073 + "escalated_frac": 0.93310546875,
3074 + "residual_disagree": 0.0
3075 + },
3076 + {
3077 + "tau": 21.375,
3078 + "escalated_frac": 0.9383138020833334,
3079 + "residual_disagree": 0.0
3080 + },
3081 + {
3082 + "tau": 21.875,
3083 + "escalated_frac": 0.9435221354166666,
3084 + "residual_disagree": 0.0
3085 + },
3086 + {
3087 + "tau": 22.25,
3088 + "escalated_frac": 0.9495442708333334,
3089 + "residual_disagree": 0.0
3090 + },
3091 + {
3092 + "tau": 22.625,
3093 + "escalated_frac": 0.9542643229166666,
3094 + "residual_disagree": 0.0
3095 + },
3096 + {
3097 + "tau": 23.130653266331706,
3098 + "escalated_frac": 0.9597981770833334,
3099 + "residual_disagree": 0.0
3100 + },
3101 + {
3102 + "tau": 23.75,
3103 + "escalated_frac": 0.9645182291666666,
3104 + "residual_disagree": 0.0
3105 + },
3106 + {
3107 + "tau": 24.375,
3108 + "escalated_frac": 0.9689127604166666,
3109 + "residual_disagree": 0.0
3110 + },
3111 + {
3112 + "tau": 25.206658291457302,
3113 + "escalated_frac": 0.9747721354166666,
3114 + "residual_disagree": 0.0
3115 + },
3116 + {
3117 + "tau": 25.875,
3118 + "escalated_frac": 0.97900390625,
3119 + "residual_disagree": 0.0
3120 + },
3121 + {
3122 + "tau": 26.875,
3123 + "escalated_frac": 0.9845377604166666,
3124 + "residual_disagree": 0.0
3125 + },
3126 + {
3127 + "tau": 28.125,
3128 + "escalated_frac": 0.9894205729166666,
3129 + "residual_disagree": 0.0
3130 + },
3131 + {
3132 + "tau": 29.766331658291392,
3133 + "escalated_frac": 0.9949544270833334,
3134 + "residual_disagree": 0.0
3135 + },
3136 + {
3137 + "tau": 35.25,
3138 + "escalated_frac": 0.9998372395833334,
3139 + "residual_disagree": 0.0
3140 + }
3141 + ],
3142 + "per_domain": {
3143 + "code": {
3144 + "agreement_rate": 0.857421875,
3145 + "auroc": 0.8844041563952945
3146 + },
3147 + "math": {
3148 + "agreement_rate": 0.9365234375,
3149 + "auroc": 0.9211117349803482
3150 + },
3151 + "chat": {
3152 + "agreement_rate": 0.8271484375,
3153 + "auroc": 0.880055229824105
3154 + },
3155 + "french": {
3156 + "agreement_rate": 0.87109375,
3157 + "auroc": 0.9024663677130045
3158 + },
3159 + "english": {
3160 + "agreement_rate": 0.84765625,
3161 + "auroc": 0.9031445704832801
3162 + },
3163 + "reasoning": {
3164 + "agreement_rate": 0.904296875,
3165 + "auroc": 0.904030942830696
3166 + }
3167 + },
3168 + "escalation_frac_for_99pct": 0.3658854166666667
3169 + },
3170 + "8": {
3171 + "bits": 8,
3172 + "n_positions": 6144,
3173 + "agreement_rate": 0.9803059895833334,
3174 + "mean_kl_ref_q": 0.003948633559048176,
3175 + "auroc_margin_predicts_disagreement": 0.9745905159697743,
3176 + "median_margin_agree": 6.25,
3177 + "median_margin_disagree": 0.25,
3178 + "escalation_curve": [
3179 + {
3180 + "tau": 0.0,
3181 + "escalated_frac": 0.0,
3182 + "residual_disagree": 0.019694010416666668
3183 + },
3184 + {
3185 + "tau": 0.0,
3186 + "escalated_frac": 0.0,
3187 + "residual_disagree": 0.019694010416666668
3188 + },
3189 + {
3190 + "tau": 0.0,
3191 + "escalated_frac": 0.0,
3192 + "residual_disagree": 0.019694010416666668
3193 + },
3194 + {
3195 + "tau": 0.125,
3196 + "escalated_frac": 0.012858072916666666,
3197 + "residual_disagree": 0.014973958333333334
3198 + },
3199 + {
3200 + "tau": 0.125,
3201 + "escalated_frac": 0.012858072916666666,
3202 + "residual_disagree": 0.014973958333333334
3203 + },
3204 + {
3205 + "tau": 0.25,
3206 + "escalated_frac": 0.022623697916666668,
3207 + "residual_disagree": 0.011393229166666666
3208 + },
3209 + {
3210 + "tau": 0.25,
3211 + "escalated_frac": 0.022623697916666668,
3212 + "residual_disagree": 0.011393229166666666
3213 + },
3214 + {
3215 + "tau": 0.25,
3216 + "escalated_frac": 0.022623697916666668,
3217 + "residual_disagree": 0.011393229166666666
3218 + },
3219 + {
3220 + "tau": 0.25,
3221 + "escalated_frac": 0.022623697916666668,
3222 + "residual_disagree": 0.011393229166666666
3223 + },
3224 + {
3225 + "tau": 0.25,
3226 + "escalated_frac": 0.022623697916666668,
3227 + "residual_disagree": 0.011393229166666666
3228 + },
3229 + {
3230 + "tau": 0.25,
3231 + "escalated_frac": 0.022623697916666668,
3232 + "residual_disagree": 0.011393229166666666
3233 + },
3234 + {
3235 + "tau": 0.375,
3236 + "escalated_frac": 0.051106770833333336,
3237 + "residual_disagree": 0.0037434895833333335
3238 + },
3239 + {
3240 + "tau": 0.5,
3241 + "escalated_frac": 0.059895833333333336,
3242 + "residual_disagree": 0.0027669270833333335
3243 + },
3244 + {
3245 + "tau": 0.5,
3246 + "escalated_frac": 0.059895833333333336,
3247 + "residual_disagree": 0.0027669270833333335
3248 + },
3249 + {
3250 + "tau": 0.5,
3251 + "escalated_frac": 0.059895833333333336,
3252 + "residual_disagree": 0.0027669270833333335
3253 + },
3254 + {
3255 + "tau": 0.5,
3256 + "escalated_frac": 0.059895833333333336,
3257 + "residual_disagree": 0.0027669270833333335
3258 + },
3259 + {
3260 + "tau": 0.5,
3261 + "escalated_frac": 0.059895833333333336,
3262 + "residual_disagree": 0.0027669270833333335
3263 + },
3264 + {
3265 + "tau": 0.625,
3266 + "escalated_frac": 0.08154296875,
3267 + "residual_disagree": 0.0011393229166666667
3268 + },
3269 + {
3270 + "tau": 0.75,
3271 + "escalated_frac": 0.08935546875,
3272 + "residual_disagree": 0.0011393229166666667
3273 + },
3274 + {
3275 + "tau": 0.75,
3276 + "escalated_frac": 0.08935546875,
3277 + "residual_disagree": 0.0011393229166666667
3278 + },
3279 + {
3280 + "tau": 0.75,
3281 + "escalated_frac": 0.08935546875,
3282 + "residual_disagree": 0.0011393229166666667
3283 + },
3284 + {
3285 + "tau": 0.75,
3286 + "escalated_frac": 0.08935546875,
3287 + "residual_disagree": 0.0011393229166666667
3288 + },
3289 + {
3290 + "tau": 0.75,
3291 + "escalated_frac": 0.08935546875,
3292 + "residual_disagree": 0.0011393229166666667
3293 + },
3294 + {
3295 + "tau": 0.875,
3296 + "escalated_frac": 0.1142578125,
3297 + "residual_disagree": 0.00048828125
3298 + },
3299 + {
3300 + "tau": 0.875,
3301 + "escalated_frac": 0.1142578125,
3302 + "residual_disagree": 0.00048828125
3303 + },
3304 + {
3305 + "tau": 1.0,
3306 + "escalated_frac": 0.12109375,
3307 + "residual_disagree": 0.00048828125
3308 + },
3309 + {
3310 + "tau": 1.0,
3311 + "escalated_frac": 0.12109375,
3312 + "residual_disagree": 0.00048828125
3313 + },
3314 + {
3315 + "tau": 1.0,
3316 + "escalated_frac": 0.12109375,
3317 + "residual_disagree": 0.00048828125
3318 + },
3319 + {
3320 + "tau": 1.0,
3321 + "escalated_frac": 0.12109375,
3322 + "residual_disagree": 0.00048828125
3323 + },
3324 + {
3325 + "tau": 1.125,
3326 + "escalated_frac": 0.14111328125,
3327 + "residual_disagree": 0.0003255208333333333
3328 + },
3329 + {
3330 + "tau": 1.25,
3331 + "escalated_frac": 0.14713541666666666,
3332 + "residual_disagree": 0.0003255208333333333
3333 + },
3334 + {
3335 + "tau": 1.25,
3336 + "escalated_frac": 0.14713541666666666,
3337 + "residual_disagree": 0.0003255208333333333
3338 + },
3339 + {
3340 + "tau": 1.25,
3341 + "escalated_frac": 0.14713541666666666,
3342 + "residual_disagree": 0.0003255208333333333
3343 + },
3344 + {
3345 + "tau": 1.25,
3346 + "escalated_frac": 0.14713541666666666,
3347 + "residual_disagree": 0.0003255208333333333
3348 + },
3349 + {
3350 + "tau": 1.375,
3351 + "escalated_frac": 0.17041015625,
3352 + "residual_disagree": 0.0
3353 + },
3354 + {
3355 + "tau": 1.5,
3356 + "escalated_frac": 0.1748046875,
3357 + "residual_disagree": 0.0
3358 + },
3359 + {
3360 + "tau": 1.5,
3361 + "escalated_frac": 0.1748046875,
3362 + "residual_disagree": 0.0
3363 + },
3364 + {
3365 + "tau": 1.5,
3366 + "escalated_frac": 0.1748046875,
3367 + "residual_disagree": 0.0
3368 + },
3369 + {
3370 + "tau": 1.5,
3371 + "escalated_frac": 0.1748046875,
3372 + "residual_disagree": 0.0
3373 + },
3374 + {
3375 + "tau": 1.625,
3376 + "escalated_frac": 0.19547526041666666,
3377 + "residual_disagree": 0.0
3378 + },
3379 + {
3380 + "tau": 1.75,
3381 + "escalated_frac": 0.20084635416666666,
3382 + "residual_disagree": 0.0
3383 + },
3384 + {
3385 + "tau": 1.75,
3386 + "escalated_frac": 0.20084635416666666,
3387 + "residual_disagree": 0.0
3388 + },
3389 + {
3390 + "tau": 1.75,
3391 + "escalated_frac": 0.20084635416666666,
3392 + "residual_disagree": 0.0
3393 + },
3394 + {
3395 + "tau": 1.75,
3396 + "escalated_frac": 0.20084635416666666,
3397 + "residual_disagree": 0.0
3398 + },
3399 + {
3400 + "tau": 1.75,
3401 + "escalated_frac": 0.20084635416666666,
3402 + "residual_disagree": 0.0
3403 + },
3404 + {
3405 + "tau": 1.875,
3406 + "escalated_frac": 0.22233072916666666,
3407 + "residual_disagree": 0.0
3408 + },
3409 + {
3410 + "tau": 2.0,
3411 + "escalated_frac": 0.22672526041666666,
3412 + "residual_disagree": 0.0
3413 + },
3414 + {
3415 + "tau": 2.0,
3416 + "escalated_frac": 0.22672526041666666,
3417 + "residual_disagree": 0.0
3418 + },
3419 + {
3420 + "tau": 2.0,
3421 + "escalated_frac": 0.22672526041666666,
3422 + "residual_disagree": 0.0
3423 + },
3424 + {
3425 + "tau": 2.125,
3426 + "escalated_frac": 0.2431640625,
3427 + "residual_disagree": 0.0
3428 + },
3429 + {
3430 + "tau": 2.25,
3431 + "escalated_frac": 0.24755859375,
3432 + "residual_disagree": 0.0
3433 + },
3434 + {
3435 + "tau": 2.25,
3436 + "escalated_frac": 0.24755859375,
3437 + "residual_disagree": 0.0
3438 + },
3439 + {
3440 + "tau": 2.25,
3441 + "escalated_frac": 0.24755859375,
3442 + "residual_disagree": 0.0
3443 + },
3444 + {
3445 + "tau": 2.375,
3446 + "escalated_frac": 0.2644856770833333,
3447 + "residual_disagree": 0.0
3448 + },
3449 + {
3450 + "tau": 2.5,
3451 + "escalated_frac": 0.2696940104166667,
3452 + "residual_disagree": 0.0
3453 + },
3454 + {
3455 + "tau": 2.5,
3456 + "escalated_frac": 0.2696940104166667,
3457 + "residual_disagree": 0.0
3458 + },
3459 + {
3460 + "tau": 2.5,
3461 + "escalated_frac": 0.2696940104166667,
3462 + "residual_disagree": 0.0
3463 + },
3464 + {
3465 + "tau": 2.5,
3466 + "escalated_frac": 0.2696940104166667,
3467 + "residual_disagree": 0.0
3468 + },
3469 + {
3470 + "tau": 2.625,
3471 + "escalated_frac": 0.2872721354166667,
3472 + "residual_disagree": 0.0
3473 + },
3474 + {
3475 + "tau": 2.75,
3476 + "escalated_frac": 0.2931315104166667,
3477 + "residual_disagree": 0.0
3478 + },
3479 + {
3480 + "tau": 2.75,
3481 + "escalated_frac": 0.2931315104166667,
3482 + "residual_disagree": 0.0
3483 + },
3484 + {
3485 + "tau": 2.75,
3486 + "escalated_frac": 0.2931315104166667,
3487 + "residual_disagree": 0.0
3488 + },
3489 + {
3490 + "tau": 2.875,
3491 + "escalated_frac": 0.3089192708333333,
3492 + "residual_disagree": 0.0
3493 + },
3494 + {
3495 + "tau": 3.0,
3496 + "escalated_frac": 0.3136393229166667,
3497 + "residual_disagree": 0.0
3498 + },
3499 + {
3500 + "tau": 3.0,
3501 + "escalated_frac": 0.3136393229166667,
3502 + "residual_disagree": 0.0
3503 + },
3504 + {
3505 + "tau": 3.0,
3506 + "escalated_frac": 0.3136393229166667,
3507 + "residual_disagree": 0.0
3508 + },
3509 + {
3510 + "tau": 3.125,
3511 + "escalated_frac": 0.3304036458333333,
3512 + "residual_disagree": 0.0
3513 + },
3514 + {
3515 + "tau": 3.25,
3516 + "escalated_frac": 0.33447265625,
3517 + "residual_disagree": 0.0
3518 + },
3519 + {
3520 + "tau": 3.25,
3521 + "escalated_frac": 0.33447265625,
3522 + "residual_disagree": 0.0
3523 + },
3524 + {
3525 + "tau": 3.25,
3526 + "escalated_frac": 0.33447265625,
3527 + "residual_disagree": 0.0
3528 + },
3529 + {
3530 + "tau": 3.375,
3531 + "escalated_frac": 0.3504231770833333,
3532 + "residual_disagree": 0.0
3533 + },
3534 + {
3535 + "tau": 3.5,
3536 + "escalated_frac": 0.35400390625,
3537 + "residual_disagree": 0.0
3538 + },
3539 + {
3540 + "tau": 3.5,
3541 + "escalated_frac": 0.35400390625,
3542 + "residual_disagree": 0.0
3543 + },
3544 + {
3545 + "tau": 3.5,
3546 + "escalated_frac": 0.35400390625,
3547 + "residual_disagree": 0.0
3548 + },
3549 + {
3550 + "tau": 3.75,
3551 + "escalated_frac": 0.37158203125,
3552 + "residual_disagree": 0.0
3553 + },
3554 + {
3555 + "tau": 3.75,
3556 + "escalated_frac": 0.37158203125,
3557 + "residual_disagree": 0.0
3558 + },
3559 + {
3560 + "tau": 3.75,
3561 + "escalated_frac": 0.37158203125,
3562 + "residual_disagree": 0.0
3563 + },
3564 + {
3565 + "tau": 3.75,
3566 + "escalated_frac": 0.37158203125,
3567 + "residual_disagree": 0.0
3568 + },
3569 + {
3570 + "tau": 3.875,
3571 + "escalated_frac": 0.3888346354166667,
3572 + "residual_disagree": 0.0
3573 + },
3574 + {
3575 + "tau": 4.0,
3576 + "escalated_frac": 0.3929036458333333,
3577 + "residual_disagree": 0.0
3578 + },
3579 + {
3580 + "tau": 4.0,
3581 + "escalated_frac": 0.3929036458333333,
3582 + "residual_disagree": 0.0
3583 + },
3584 + {
3585 + "tau": 4.177135678391949,
3586 + "escalated_frac": 0.4070638020833333,
3587 + "residual_disagree": 0.0
3588 + },
3589 + {
3590 + "tau": 4.25,
3591 + "escalated_frac": 0.4070638020833333,
3592 + "residual_disagree": 0.0
3593 + },
3594 + {
3595 + "tau": 4.25,
3596 + "escalated_frac": 0.4070638020833333,
3597 + "residual_disagree": 0.0
3598 + },
3599 + {
3600 + "tau": 4.375,
3601 + "escalated_frac": 0.42138671875,
3602 + "residual_disagree": 0.0
3603 + },
3604 + {
3605 + "tau": 4.5,
3606 + "escalated_frac": 0.42529296875,
3607 + "residual_disagree": 0.0
3608 + },
3609 + {
3610 + "tau": 4.5,
3611 + "escalated_frac": 0.42529296875,
3612 + "residual_disagree": 0.0
3613 + },
3614 + {
3615 + "tau": 4.625,
3616 + "escalated_frac": 0.4358723958333333,
3617 + "residual_disagree": 0.0
3618 + },
3619 + {
3620 + "tau": 4.75,
3621 + "escalated_frac": 0.43896484375,
3622 + "residual_disagree": 0.0
3623 + },
3624 + {
3625 + "tau": 4.75,
3626 + "escalated_frac": 0.43896484375,
3627 + "residual_disagree": 0.0
3628 + },
3629 + {
3630 + "tau": 4.875,
3631 + "escalated_frac": 0.4503580729166667,
3632 + "residual_disagree": 0.0
3633 + },
3634 + {
3635 + "tau": 5.0,
3636 + "escalated_frac": 0.4529622395833333,
3637 + "residual_disagree": 0.0
3638 + },
3639 + {
3640 + "tau": 5.0,
3641 + "escalated_frac": 0.4529622395833333,
3642 + "residual_disagree": 0.0
3643 + },
3644 + {
3645 + "tau": 5.25,
3646 + "escalated_frac": 0.4658203125,
3647 + "residual_disagree": 0.0
3648 + },
3649 + {
3650 + "tau": 5.25,
3651 + "escalated_frac": 0.4658203125,
3652 + "residual_disagree": 0.0
3653 + },
3654 + {
3655 + "tau": 5.375,
3656 + "escalated_frac": 0.4752604166666667,
3657 + "residual_disagree": 0.0
3658 + },
3659 + {
3660 + "tau": 5.5,
3661 + "escalated_frac": 0.4775390625,
3662 + "residual_disagree": 0.0
3663 + },
3664 + {
3665 + "tau": 5.625,
3666 + "escalated_frac": 0.4871419270833333,
3667 + "residual_disagree": 0.0
3668 + },
3669 + {
3670 + "tau": 5.75,
3671 + "escalated_frac": 0.4900716145833333,
3672 + "residual_disagree": 0.0
3673 + },
3674 + {
3675 + "tau": 5.875,
3676 + "escalated_frac": 0.4972330729166667,
3677 + "residual_disagree": 0.0
3678 + },
3679 + {
3680 + "tau": 6.0,
3681 + "escalated_frac": 0.49951171875,
3682 + "residual_disagree": 0.0
3683 + },
3684 + {
3685 + "tau": 6.125,
3686 + "escalated_frac": 0.5066731770833334,
3687 + "residual_disagree": 0.0
3688 + },
3689 + {
3690 + "tau": 6.25,
3691 + "escalated_frac": 0.50927734375,
3692 + "residual_disagree": 0.0
3693 + },
3694 + {
3695 + "tau": 6.25,
3696 + "escalated_frac": 0.50927734375,
3697 + "residual_disagree": 0.0
3698 + },
3699 + {
3700 + "tau": 6.5,
3701 + "escalated_frac": 0.52099609375,
3702 + "residual_disagree": 0.0
3703 + },
3704 + {
3705 + "tau": 6.5,
3706 + "escalated_frac": 0.52099609375,
3707 + "residual_disagree": 0.0
3708 + },
3709 + {
3710 + "tau": 6.75,
3711 + "escalated_frac": 0.53173828125,
3712 + "residual_disagree": 0.0
3713 + },
3714 + {
3715 + "tau": 6.75,
3716 + "escalated_frac": 0.53173828125,
3717 + "residual_disagree": 0.0
3718 + },
3719 + {
3720 + "tau": 6.875,
3721 + "escalated_frac": 0.541015625,
3722 + "residual_disagree": 0.0
3723 + },
3724 + {
3725 + "tau": 7.0,
3726 + "escalated_frac": 0.54296875,
3727 + "residual_disagree": 0.0
3728 + },
3729 + {
3730 + "tau": 7.125,
3731 + "escalated_frac": 0.5499674479166666,
3732 + "residual_disagree": 0.0
3733 + },
3734 + {
3735 + "tau": 7.25,
3736 + "escalated_frac": 0.5540364583333334,
3737 + "residual_disagree": 0.0
3738 + },
3739 + {
3740 + "tau": 7.5,
3741 + "escalated_frac": 0.5626627604166666,
3742 + "residual_disagree": 0.0
3743 + },
3744 + {
3745 + "tau": 7.5,
3746 + "escalated_frac": 0.5626627604166666,
3747 + "residual_disagree": 0.0
3748 + },
3749 + {
3750 + "tau": 7.75,
3751 + "escalated_frac": 0.5709635416666666,
3752 + "residual_disagree": 0.0
3753 + },
3754 + {
3755 + "tau": 7.75,
3756 + "escalated_frac": 0.5709635416666666,
3757 + "residual_disagree": 0.0
3758 + },
3759 + {
3760 + "tau": 8.0,
3761 + "escalated_frac": 0.5802408854166666,
3762 + "residual_disagree": 0.0
3763 + },
3764 + {
3765 + "tau": 8.125,
3766 + "escalated_frac": 0.5852864583333334,
3767 + "residual_disagree": 0.0
3768 + },
3769 + {
3770 + "tau": 8.25,
3771 + "escalated_frac": 0.5880533854166666,
3772 + "residual_disagree": 0.0
3773 + },
3774 + {
3775 + "tau": 8.5,
3776 + "escalated_frac": 0.59765625,
3777 + "residual_disagree": 0.0
3778 + },
3779 + {
3780 + "tau": 8.5,
3781 + "escalated_frac": 0.59765625,
3782 + "residual_disagree": 0.0
3783 + },
3784 + {
3785 + "tau": 8.648869346733647,
3786 + "escalated_frac": 0.6080729166666666,
3787 + "residual_disagree": 0.0
3788 + },
3789 + {
3790 + "tau": 8.75,
3791 + "escalated_frac": 0.6080729166666666,
3792 + "residual_disagree": 0.0
3793 + },
3794 + {
3795 + "tau": 9.0,
3796 + "escalated_frac": 0.6160481770833334,
3797 + "residual_disagree": 0.0
3798 + },
3799 + {
3800 + "tau": 9.125,
3801 + "escalated_frac": 0.6222330729166666,
3802 + "residual_disagree": 0.0
3803 + },
3804 + {
3805 + "tau": 9.25,
3806 + "escalated_frac": 0.6253255208333334,
3807 + "residual_disagree": 0.0
3808 + },
3809 + {
3810 + "tau": 9.375,
3811 + "escalated_frac": 0.6326497395833334,
3812 + "residual_disagree": 0.0
3813 + },
3814 + {
3815 + "tau": 9.5,
3816 + "escalated_frac": 0.63623046875,
3817 + "residual_disagree": 0.0
3818 + },
3819 + {
3820 + "tau": 9.625,
3821 + "escalated_frac": 0.6416015625,
3822 + "residual_disagree": 0.0
3823 + },
3824 + {
3825 + "tau": 9.75,
3826 + "escalated_frac": 0.6433919270833334,
3827 + "residual_disagree": 0.0
3828 + },
3829 + {
3830 + "tau": 10.0,
3831 + "escalated_frac": 0.6531575520833334,
3832 + "residual_disagree": 0.0
3833 + },
3834 + {
3835 + "tau": 10.0,
3836 + "escalated_frac": 0.6531575520833334,
3837 + "residual_disagree": 0.0
3838 + },
3839 + {
3840 + "tau": 10.25,
3841 + "escalated_frac": 0.66064453125,
3842 + "residual_disagree": 0.0
3843 + },
3844 + {
3845 + "tau": 10.5,
3846 + "escalated_frac": 0.6673177083333334,
3847 + "residual_disagree": 0.0
3848 + },
3849 + {
3850 + "tau": 10.75,
3851 + "escalated_frac": 0.6726888020833334,
3852 + "residual_disagree": 0.0
3853 + },
3854 + {
3855 + "tau": 10.75,
3856 + "escalated_frac": 0.6726888020833334,
3857 + "residual_disagree": 0.0
3858 + },
3859 + {
3860 + "tau": 11.0,
3861 + "escalated_frac": 0.681640625,
3862 + "residual_disagree": 0.0
3863 + },
3864 + {
3865 + "tau": 11.25,
3866 + "escalated_frac": 0.6883138020833334,
3867 + "residual_disagree": 0.0
3868 + },
3869 + {
3870 + "tau": 11.3712311557789,
3871 + "escalated_frac": 0.693359375,
3872 + "residual_disagree": 0.0
3873 + },
3874 + {
3875 + "tau": 11.5,
3876 + "escalated_frac": 0.6954752604166666,
3877 + "residual_disagree": 0.0
3878 + },
3879 + {
3880 + "tau": 11.75,
3881 + "escalated_frac": 0.7021484375,
3882 + "residual_disagree": 0.0
3883 + },
3884 + {
3885 + "tau": 11.875,
3886 + "escalated_frac": 0.70556640625,
3887 + "residual_disagree": 0.0
3888 + },
3889 + {
3890 + "tau": 12.0,
3891 + "escalated_frac": 0.708984375,
3892 + "residual_disagree": 0.0
3893 + },
3894 + {
3895 + "tau": 12.25,
3896 + "escalated_frac": 0.71728515625,
3897 + "residual_disagree": 0.0
3898 + },
3899 + {
3900 + "tau": 12.375,
3901 + "escalated_frac": 0.7205403645833334,
3902 + "residual_disagree": 0.0
3903 + },
3904 + {
3905 + "tau": 12.625,
3906 + "escalated_frac": 0.7278645833333334,
3907 + "residual_disagree": 0.0
3908 + },
3909 + {
3910 + "tau": 12.75,
3911 + "escalated_frac": 0.7301432291666666,
3912 + "residual_disagree": 0.0
3913 + },
3914 + {
3915 + "tau": 13.0,
3916 + "escalated_frac": 0.7371419270833334,
3917 + "residual_disagree": 0.0
3918 + },
3919 + {
3920 + "tau": 13.25,
3921 + "escalated_frac": 0.7425130208333334,
3922 + "residual_disagree": 0.0
3923 + },
3924 + {
3925 + "tau": 13.5,
3926 + "escalated_frac": 0.7478841145833334,
3927 + "residual_disagree": 0.0
3928 + },
3929 + {
3930 + "tau": 13.75,
3931 + "escalated_frac": 0.7527669270833334,
3932 + "residual_disagree": 0.0
3933 + },
3934 + {
3935 + "tau": 14.0,
3936 + "escalated_frac": 0.7584635416666666,
3937 + "residual_disagree": 0.0
3938 + },
3939 + {
3940 + "tau": 14.125,
3941 + "escalated_frac": 0.7633463541666666,
3942 + "residual_disagree": 0.0
3943 + },
3944 + {
3945 + "tau": 14.25,
3946 + "escalated_frac": 0.76611328125,
3947 + "residual_disagree": 0.0
3948 + },
3949 + {
3950 + "tau": 14.5,
3951 + "escalated_frac": 0.7716471354166666,
3952 + "residual_disagree": 0.0
3953 + },
3954 + {
3955 + "tau": 14.75,
3956 + "escalated_frac": 0.7786458333333334,
3957 + "residual_disagree": 0.0
3958 + },
3959 + {
3960 + "tau": 14.875,
3961 + "escalated_frac": 0.7830403645833334,
3962 + "residual_disagree": 0.0
3963 + },
3964 + {
3965 + "tau": 15.0,
3966 + "escalated_frac": 0.7867838541666666,
3967 + "residual_disagree": 0.0
3968 + },
3969 + {
3970 + "tau": 15.25,
3971 + "escalated_frac": 0.7937825520833334,
3972 + "residual_disagree": 0.0
3973 + },
3974 + {
3975 + "tau": 15.375,
3976 + "escalated_frac": 0.796875,
3977 + "residual_disagree": 0.0
3978 + },
3979 + {
3980 + "tau": 15.5,
3981 + "escalated_frac": 0.8009440104166666,
3982 + "residual_disagree": 0.0
3983 + },
3984 + {
3985 + "tau": 15.75,
3986 + "escalated_frac": 0.8082682291666666,
3987 + "residual_disagree": 0.0
3988 + },
3989 + {
3990 + "tau": 15.875,
3991 + "escalated_frac": 0.8136393229166666,
3992 + "residual_disagree": 0.0
3993 + },
3994 + {
3995 + "tau": 16.087939698492505,
3996 + "escalated_frac": 0.8190104166666666,
3997 + "residual_disagree": 0.0
3998 + },
3999 + {
4000 + "tau": 16.375,
4001 + "escalated_frac": 0.8230794270833334,
4002 + "residual_disagree": 0.0
4003 + },
4004 + {
4005 + "tau": 16.625,
4006 + "escalated_frac": 0.828125,
4007 + "residual_disagree": 0.0
4008 + },
4009 + {
4010 + "tau": 16.75,
4011 + "escalated_frac": 0.8312174479166666,
4012 + "residual_disagree": 0.0
4013 + },
4014 + {
4015 + "tau": 17.0,
4016 + "escalated_frac": 0.8380533854166666,
4017 + "residual_disagree": 0.0
4018 + },
4019 + {
4020 + "tau": 17.25,
4021 + "escalated_frac": 0.8435872395833334,
4022 + "residual_disagree": 0.0
4023 + },
4024 + {
4025 + "tau": 17.5,
4026 + "escalated_frac": 0.8479817708333334,
4027 + "residual_disagree": 0.0
4028 + },
4029 + {
4030 + "tau": 17.75,
4031 + "escalated_frac": 0.8528645833333334,
4032 + "residual_disagree": 0.0
4033 + },
4034 + {
4035 + "tau": 18.0,
4036 + "escalated_frac": 0.8575846354166666,
4037 + "residual_disagree": 0.0
4038 + },
4039 + {
4040 + "tau": 18.190954773869294,
4041 + "escalated_frac": 0.8642578125,
4042 + "residual_disagree": 0.0
4043 + },
4044 + {
4045 + "tau": 18.5,
4046 + "escalated_frac": 0.86767578125,
4047 + "residual_disagree": 0.0
4048 + },
4049 + {
4050 + "tau": 18.75,
4051 + "escalated_frac": 0.8733723958333334,
4052 + "residual_disagree": 0.0
4053 + },
4054 + {
4055 + "tau": 19.125,
4056 + "escalated_frac": 0.8790690104166666,
4057 + "residual_disagree": 0.0
4058 + },
4059 + {
4060 + "tau": 19.5,
4061 + "escalated_frac": 0.88427734375,
4062 + "residual_disagree": 0.0
4063 + },
4064 + {
4065 + "tau": 19.75,
4066 + "escalated_frac": 0.8885091145833334,
4067 + "residual_disagree": 0.0
4068 + },
4069 + {
4070 + "tau": 20.125,
4071 + "escalated_frac": 0.8935546875,
4072 + "residual_disagree": 0.0
4073 + },
4074 + {
4075 + "tau": 20.5,
4076 + "escalated_frac": 0.8982747395833334,
4077 + "residual_disagree": 0.0
4078 + },
4079 + {
4080 + "tau": 20.8103015075377,
4081 + "escalated_frac": 0.9044596354166666,
4082 + "residual_disagree": 0.0
4083 + },
4084 + {
4085 + "tau": 21.125,
4086 + "escalated_frac": 0.9090169270833334,
4087 + "residual_disagree": 0.0
4088 + },
4089 + {
4090 + "tau": 21.5,
4091 + "escalated_frac": 0.91357421875,
4092 + "residual_disagree": 0.0
4093 + },
4094 + {
4095 + "tau": 21.875,
4096 + "escalated_frac": 0.9182942708333334,
4097 + "residual_disagree": 0.0
4098 + },
4099 + {
4100 + "tau": 22.25,
4101 + "escalated_frac": 0.9225260416666666,
4102 + "residual_disagree": 0.0
4103 + },
4104 + {
4105 + "tau": 22.625,
4106 + "escalated_frac": 0.9285481770833334,
4107 + "residual_disagree": 0.0
4108 + },
4109 + {
4110 + "tau": 23.0,
4111 + "escalated_frac": 0.93310546875,
4112 + "residual_disagree": 0.0
4113 + },
4114 + {
4115 + "tau": 23.5,
4116 + "escalated_frac": 0.9391276041666666,
4117 + "residual_disagree": 0.0
4118 + },
4119 + {
4120 + "tau": 23.875,
4121 + "escalated_frac": 0.9436848958333334,
4122 + "residual_disagree": 0.0
4123 + },
4124 + {
4125 + "tau": 24.375,
4126 + "escalated_frac": 0.9495442708333334,
4127 + "residual_disagree": 0.0
4128 + },
4129 + {
4130 + "tau": 24.88599246231155,
4131 + "escalated_frac": 0.9547526041666666,
4132 + "residual_disagree": 0.0
4133 + },
4134 + {
4135 + "tau": 25.25,
4136 + "escalated_frac": 0.9578450520833334,
4137 + "residual_disagree": 0.0
4138 + },
4139 + {
4140 + "tau": 25.875,
4141 + "escalated_frac": 0.9645182291666666,
4142 + "residual_disagree": 0.0
4143 + },
4144 + {
4145 + "tau": 26.695979899497388,
4146 + "escalated_frac": 0.9697265625,
4147 + "residual_disagree": 0.0
4148 + },
4149 + {
4150 + "tau": 27.25,
4151 + "escalated_frac": 0.9744466145833334,
4152 + "residual_disagree": 0.0
4153 + },
4154 + {
4155 + "tau": 28.25,
4156 + "escalated_frac": 0.9794921875,
4157 + "residual_disagree": 0.0
4158 + },
4159 + {
4160 + "tau": 29.5,
4161 + "escalated_frac": 0.9845377604166666,
4162 + "residual_disagree": 0.0
4163 + },
4164 + {
4165 + "tau": 30.5,
4166 + "escalated_frac": 0.98974609375,
4167 + "residual_disagree": 0.0
4168 + },
4169 + {
4170 + "tau": 32.032663316582784,
4171 + "escalated_frac": 0.9949544270833334,
4172 + "residual_disagree": 0.0
4173 + },
4174 + {
4175 + "tau": 37.5,
4176 + "escalated_frac": 0.9998372395833334,
4177 + "residual_disagree": 0.0
4178 + }
4179 + ],
4180 + "per_domain": {
4181 + "code": {
4182 + "agreement_rate": 0.982421875,
4183 + "auroc": 0.9611221559531699
4184 + },
4185 + "math": {
4186 + "agreement_rate": 0.9912109375,
4187 + "auroc": 0.9871373836891079
4188 + },
4189 + "chat": {
4190 + "agreement_rate": 0.974609375,
4191 + "auroc": 0.968937875751503
4192 + },
4193 + "french": {
4194 + "agreement_rate": 0.978515625,
4195 + "auroc": 0.981854472872437
4196 + },
4197 + "english": {
4198 + "agreement_rate": 0.9677734375,
4199 + "auroc": 0.9724490107941167
4200 + },
4201 + "reasoning": {
4202 + "agreement_rate": 0.9873046875,
4203 + "auroc": 0.9800654340713688
4204 + }
4205 + },
4206 + "escalation_frac_for_99pct": 0.051106770833333336
4207 + }
4208 + }
4209 +}
\ No newline at end of file
4210