Freeze confirmatory design (manifest, registry stamp) and register the 288-run assignment (seed 301)
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
7 changed files +11,234 −24
modified
protocol/amendments.csv
+1 −0
@@ -1,3 +1,4 @@ | |||||
| 1 | amendment_id,timestamp_utc,phase,before_or_after_results,document,description,rationale,previous_version_ref | 1 | amendment_id,timestamp_utc,phase,before_or_after_results,document,description,rationale,previous_version_ref | ||
| 2 | A-001,2026-10-30T03:00:00Z,02,before_results,configs/budgets.yaml,"Provisional live-inference envelope set by the executor: per benchmark run 4.00 USD / 1500 s / 80 turns per session; per transfer run 12.00 USD / 3600 s; total study envelope 800 USD (pilot + main + transfer + recovery + ablations) with the confirmatory scale to be chosen after the pilot.","The mandate requires a precise budget decision but none was stated; the researcher is not available mid-run; prior UQO studies spent ≈1,079 USD of API inference (WP26), so 800 USD is a conservative envelope. The researcher may raise or lower it; any change will be a new amendment.",n/a | 2 | A-001,2026-10-30T03:00:00Z,02,before_results,configs/budgets.yaml,"Provisional live-inference envelope set by the executor: per benchmark run 4.00 USD / 1500 s / 80 turns per session; per transfer run 12.00 USD / 3600 s; total study envelope 800 USD (pilot + main + transfer + recovery + ablations) with the confirmatory scale to be chosen after the pilot.","The mandate requires a precise budget decision but none was stated; the researcher is not available mid-run; prior UQO studies spent ≈1,079 USD of API inference (WP26), so 800 USD is a conservative envelope. The researcher may raise or lower it; any change will be a new amendment.",n/a | ||
| 3 | A-002,2026-10-30T04:40:00Z,06,before_confirmatory_results,src/pgar/evaluation.py,"Final evaluators are now executed on a composed workspace (public bundle + the run's collected out/) instead of the run directory alone, which lacked data/ and made source-grounded checks (F4 'numbers_in_sources') fail for correct answers. All pilot runs evaluated before the fix are re-evaluated with the same oracles; original receipts are kept in runs/pilot/<run>/evaluation_receipt_pre-A002.json.","Pilot diagnosis: F4-dev-01 failed in all four arms with a spurious fabrication violation although every required criterion passed. Harness defect, not an oracle change.",runs/pilot/*/evaluation_receipt_pre-A002.json | 3 | A-002,2026-10-30T04:40:00Z,06,before_confirmatory_results,src/pgar/evaluation.py,"Final evaluators are now executed on a composed workspace (public bundle + the run's collected out/) instead of the run directory alone, which lacked data/ and made source-grounded checks (F4 'numbers_in_sources') fail for correct answers. All pilot runs evaluated before the fix are re-evaluated with the same oracles; original receipts are kept in runs/pilot/<run>/evaluation_receipt_pre-A002.json.","Pilot diagnosis: F4-dev-01 failed in all four arms with a spurious fabrication violation although every required criterion passed. Harness defect, not an oracle change.",runs/pilot/*/evaluation_receipt_pre-A002.json | ||
| 4 | +A-003,2026-10-30T05:40:00Z,06,before_confirmatory_results,protocol/preregistration.md,"Preregistration frozen as v1.0 (adds §10 pilot-informed decisions: confirmatory design 24 tasks × 4 arms × 3 repeats = 288 runs, seed 301); freeze manifest protocol/freeze_manifest_conf.json records SHA-256 of all 24 confirmatory bundles and oracles, arm texts, gate, launcher, scheduler, evaluator, configs and analysis scripts; tasks/registry.csv stamped frozen_utc.","Scale decision per §6 of the preregistration using measured pilot costs; no hypothesis, endpoint or analysis rule changed.",git 2c82656 | ||||
added
protocol/freeze_manifest_conf.json
+263 −0
@@ -0,0 +1,263 @@ | |||||
| 1 | +{ | ||||
| 2 | + "frozen_utc": "2026-10-30T04:02:39+00:00", | ||||
| 3 | + "split": "conf", | ||||
| 4 | + "bundles": { | ||||
| 5 | + "F1-conf-01": { | ||||
| 6 | + "bundle_json": "a61750dcacf0ee144e2f1612a757cf3b77d74ff0c8a4d05f2ef37fa2dbd94105", | ||||
| 7 | + "TASK_md": "a1445e54705eb87a7873503e75c51e4f37ad8594693f10839f681fb9238ca91a", | ||||
| 8 | + "gate_spec": "2435fd30a9f36563592bbfefe3c379975747affc40c42d521c3227215aaf7619" | ||||
| 9 | + }, | ||||
| 10 | + "F1-conf-02": { | ||||
| 11 | + "bundle_json": "c13a30fbc4fa5f8502813a0b7c26d62dad6f731f5dafba69c11fd09e8608f0a1", | ||||
| 12 | + "TASK_md": "979bd92bb25ae3dda2a47ef6fce427a1db4a9c3f848a86bf09c4f424fcfe6dcb", | ||||
| 13 | + "gate_spec": "5e2280ad93e5a7edb7d300c499d0e2a12c442513edfe1f25b85c24c282bc01bc" | ||||
| 14 | + }, | ||||
| 15 | + "F1-conf-03": { | ||||
| 16 | + "bundle_json": "a251c5dd2b17dad0a01466fe57338c9c9f3dfbcfed40d69d8205334364a0f1c9", | ||||
| 17 | + "TASK_md": "89b3ae62efc5d41942ef1484b256066f58b5e6075d63edcbbc69f53e76cdd438", | ||||
| 18 | + "gate_spec": "485525fd3cc180616a452c24f04aec384c433768a96eef644650ff8d5bfcc5d4" | ||||
| 19 | + }, | ||||
| 20 | + "F1-conf-04": { | ||||
| 21 | + "bundle_json": "0b9030ef4081e6e71f37c13b0009d15acdebb7df22f40be9eb4d8e68b396e17d", | ||||
| 22 | + "TASK_md": "53fa917644a6c2c67f9a5a7939fd7681fbb32b60ca39cac8d178188f76b5137e", | ||||
| 23 | + "gate_spec": "1ca2626b0090513853be28b4d434b8b27fa53841260307563418a3eaef27ff1d" | ||||
| 24 | + }, | ||||
| 25 | + "F2-conf-01": { | ||||
| 26 | + "bundle_json": "b5e6926b8612fe84faac744b008f5bfa99c6fcd32e03f4c3262162a983ee8334", | ||||
| 27 | + "TASK_md": "114f43f61885b229d7f982f413fd2d347e7fa7be0509555a18067c7cbf1b7cad", | ||||
| 28 | + "gate_spec": "9785bd6f17ce8946989d00b4f847cf13136c7bc3b33f2c7a4c1ff010c214ada4" | ||||
| 29 | + }, | ||||
| 30 | + "F2-conf-02": { | ||||
| 31 | + "bundle_json": "78aa39791388526a1be69f2f020c6bacf51ca0487d2d314341d29745c1dcd42c", | ||||
| 32 | + "TASK_md": "447a35f0911fb5fa0ee410db5519883ef0774fc71524057bdffba5a00de148f0", | ||||
| 33 | + "gate_spec": "268541a1253f33df6fdc3a6fb21b1318c1c65b8429adb1aacd1e21b050baa9c6" | ||||
| 34 | + }, | ||||
| 35 | + "F2-conf-03": { | ||||
| 36 | + "bundle_json": "1a8772c6d026ea816c2c9460a9bd61af2625f01427245672c86df5099cb78f9c", | ||||
| 37 | + "TASK_md": "62706db4c415f5ece303e1f68f949d1f9719ac2b7430d9733ed29965da8ab7b0", | ||||
| 38 | + "gate_spec": "a6c3198ab94269bf8b99585e56d88cea490ee73143a8b879bc1ac5fbfa74a1ef" | ||||
| 39 | + }, | ||||
| 40 | + "F2-conf-04": { | ||||
| 41 | + "bundle_json": "2ccad59f602537f40051fddebde3ea12b90a8547d163594df2d0420352e0339e", | ||||
| 42 | + "TASK_md": "b80f25d137f0d04f5223537e07525015a8d6257e5084a83ce87866da7a71a0eb", | ||||
| 43 | + "gate_spec": "939144536131765675ba640605191be350be3312c9098770179c2db3c9096991" | ||||
| 44 | + }, | ||||
| 45 | + "F3-conf-01": { | ||||
| 46 | + "bundle_json": "3d3666ddb8c8b19c93f6cea1ef983368f52f9eb6cee4696a19ad82111cab35b8", | ||||
| 47 | + "TASK_md": "b1385cbc4e56444c3aeb8f948062b345ed14ef346ab2fad744a9f5a6549fad0e", | ||||
| 48 | + "gate_spec": "240492f4e6d528fdf529ad5d19987f04aac9f15e90c60902f6e847967524f3f1" | ||||
| 49 | + }, | ||||
| 50 | + "F3-conf-02": { | ||||
| 51 | + "bundle_json": "72bd0760441017e0cd4b184dcd5b514b61e577d851e895d582c6a245515a2603", | ||||
| 52 | + "TASK_md": "266ce4fad79c502d3b1561b76dada677c9276f020c8315cd0915d432bfd505ed", | ||||
| 53 | + "gate_spec": "e17b24eef15436583c727006c17022b1749c3eae73afa72ea55dd5e2b81344b6" | ||||
| 54 | + }, | ||||
| 55 | + "F3-conf-03": { | ||||
| 56 | + "bundle_json": "f3e5a971802b57e92ae4fad13bebd464e6a5dc2f755df9aa9420e2b515763cf8", | ||||
| 57 | + "TASK_md": "05343e6f63af28c29a93e789bd621a175efcbd63bafd46f524d3bde9ca21dd1e", | ||||
| 58 | + "gate_spec": "7f7e7d36571218fec139ba1dbf9a3f01ee6c0bac7bc8238debd58d44a69c7071" | ||||
| 59 | + }, | ||||
| 60 | + "F3-conf-04": { | ||||
| 61 | + "bundle_json": "60423555f5292067aebaa897310813f5bbbca474e63d2256e1cb8a67bbad2a1a", | ||||
| 62 | + "TASK_md": "3039dc66364cbd8ca986e67cd619f2be53509c6cf77f55bc179583ee85464810", | ||||
| 63 | + "gate_spec": "3585c0dae94511350806b5f595e10839bd205eb59763dac49deb6a66b1396eab" | ||||
| 64 | + }, | ||||
| 65 | + "F4-conf-01": { | ||||
| 66 | + "bundle_json": "4f8e4fcee6110ddaaf15267a53dbbcdec8b1b0de7e9ec7c0db85d32ef7819b57", | ||||
| 67 | + "TASK_md": "c0ca541d98a4fc46ff5a6dd27fc6e09ae1875e5a5ce7c2a8d963f80138352d20", | ||||
| 68 | + "gate_spec": "eb563a002112a161b73ac3c4c1f83feafc9ff733f35db10c13f1c95503ea6fc5" | ||||
| 69 | + }, | ||||
| 70 | + "F4-conf-02": { | ||||
| 71 | + "bundle_json": "78a0db5fd3108905cf14cda85fdf7ce155c29c3435c3cb69a8a427a52fc6d6ec", | ||||
| 72 | + "TASK_md": "2f517ee00f69f6ba411c5189a9358a21afc763b36c88f8420603c079421a7c1a", | ||||
| 73 | + "gate_spec": "c7b67c16ad001e747b68be8a277a0e635f35013ccc3d44ef0f0a6270ec28991c" | ||||
| 74 | + }, | ||||
| 75 | + "F4-conf-03": { | ||||
| 76 | + "bundle_json": "b7f53f8118e35079e33ec7f4472d09719d5331039049ca1fb073fc2d61f43d58", | ||||
| 77 | + "TASK_md": "6196aad6cf54ad79a83715ebb1c68e08c9fc03b96bceeeb868e7a9785f839b0c", | ||||
| 78 | + "gate_spec": "6525a12ead9042fbe3481cef51038c3bb85e39aea33c7b4102e07a13fb704804" | ||||
| 79 | + }, | ||||
| 80 | + "F4-conf-04": { | ||||
| 81 | + "bundle_json": "73626fc9918061fb09d68315f0b5c65d88db269da2c666b9f24860f2bc1b53e5", | ||||
| 82 | + "TASK_md": "ff798455d4e67f177f3109ee648dea2581441bbf14dcf66217357b97142df2a8", | ||||
| 83 | + "gate_spec": "2333dfa3f0ef0db6a35419a41b6a696ac5b4547b7b9eee208af28f9507868392" | ||||
| 84 | + }, | ||||
| 85 | + "F5-conf-01": { | ||||
| 86 | + "bundle_json": "18e23787d706db137b2a177d24890cc63194af45df5bfae781bb0f16df09c214", | ||||
| 87 | + "TASK_md": "2c56e5f2c0baba2af82ce4e42e2a9ca546f980c80b64532ccb4fba710eacf9a4", | ||||
| 88 | + "gate_spec": "48493d7327c718a496ab94921c6e9e5a9be62359b9f315537cdf76859e35f088" | ||||
| 89 | + }, | ||||
| 90 | + "F5-conf-02": { | ||||
| 91 | + "bundle_json": "cbad893d5aff4b816b6c85200a78740ae10635b39d255672b74735d8c813a022", | ||||
| 92 | + "TASK_md": "2a389cdf509f84a0f7e0cf2780400eb5b6c1cd45ee0359a2d0708b803641d090", | ||||
| 93 | + "gate_spec": "1479a11a8587f31338a07b4532d5b9c2da1bba66c2486348acebf18090c30c6b" | ||||
| 94 | + }, | ||||
| 95 | + "F5-conf-03": { | ||||
| 96 | + "bundle_json": "2304c041230e67ebc977b7ad780dea3097960e770bcbedee9edb6cc3078c861a", | ||||
| 97 | + "TASK_md": "31fbaa2d88530fab7509ed52d12759f5b4fc91be27ba902545f08663f9c431d6", | ||||
| 98 | + "gate_spec": "fe547a7a065d89426b9ef65fd59a96354191a660f0b1ce99e812e26e54b02c5d" | ||||
| 99 | + }, | ||||
| 100 | + "F5-conf-04": { | ||||
| 101 | + "bundle_json": "69a692e77615b09bf2128186ae89047d1dd52f8608d41753af6ef7a7c3084209", | ||||
| 102 | + "TASK_md": "7723e3dc76d5ee4831121d96c79e04c503d90569ba31c5a92d4c31bb24f0f3d2", | ||||
| 103 | + "gate_spec": "857972db0a035afd491bcfe1b5dc15095377a1fa6374a474f0e37cf424fb3860" | ||||
| 104 | + }, | ||||
| 105 | + "F6-conf-01": { | ||||
| 106 | + "bundle_json": "05a887612a7b92dfc377c3bbb30c7eaefad25bbc4be61add470162f58244d641", | ||||
| 107 | + "TASK_md": "2283a94e1a3aa085d881cacc8bc6b793b6334b6193373788a992080c86a2d89e", | ||||
| 108 | + "gate_spec": "d53691940951460bcc45cb08c958917841ae1bfbc8d242d37ded795a3e998eb7" | ||||
| 109 | + }, | ||||
| 110 | + "F6-conf-02": { | ||||
| 111 | + "bundle_json": "f8236a8d4aff1da5d035c8c5a5a203abbefcb092e0ba36f04c6ed12b3da5da88", | ||||
| 112 | + "TASK_md": "72bd0f949b77e7176436a43304093a682256ca4b693bdbe425df71e9452472d0", | ||||
| 113 | + "gate_spec": "456ff7a4c127630bef706b95a35d172a5fa7ae71112c182f591e38dd6ecd31ba" | ||||
| 114 | + }, | ||||
| 115 | + "F6-conf-03": { | ||||
| 116 | + "bundle_json": "26ee50278611967d5a80428b531496d9c1fe94953c8b0eae4e0e15c57c96d0aa", | ||||
| 117 | + "TASK_md": "64afb4744b89e1ab9b45cefc3fad325b1857216f04a3d3cc54ed3b1681db2a0e", | ||||
| 118 | + "gate_spec": "bced13a911d728cc853d5a88f4412cf86338d82c325eb7891ba2f5fd86317f61" | ||||
| 119 | + }, | ||||
| 120 | + "F6-conf-04": { | ||||
| 121 | + "bundle_json": "fa4857dfe51956cfef267a76a87b597efe70262f616674fdf4e76ca501d8059d", | ||||
| 122 | + "TASK_md": "f91b6d78e16573314371e5761a0e79c69721af82a9454655f56a864b34006c6e", | ||||
| 123 | + "gate_spec": "7233dd8b32d0b3680b8275042673b5278441b86744044bd806094c4af34bd87f" | ||||
| 124 | + } | ||||
| 125 | + }, | ||||
| 126 | + "oracles": { | ||||
| 127 | + "F1-conf-01": { | ||||
| 128 | + "oracle_json": "58419a6c63b3fe5ec0632cfa1c81fb208c253231abd06d1b55b0568c1f168d2e", | ||||
| 129 | + "evaluate_py": "a00935bf48e126721f1e83a8e481ab28d0575716cdf3246d3cb0527eaae46fe5" | ||||
| 130 | + }, | ||||
| 131 | + "F1-conf-02": { | ||||
| 132 | + "oracle_json": "88041d60bfa1855cd036fae779e20faa54360c97a27f3bd8cce7e8f3e114e979", | ||||
| 133 | + "evaluate_py": "3b1a5901b88374c0da61f83822e6f4b0091fe713115a1f752583c8fec1fc3832" | ||||
| 134 | + }, | ||||
| 135 | + "F1-conf-03": { | ||||
| 136 | + "oracle_json": "da310f306e465ff728eff6dcedb3dd0ee14496cc8e05643b02047fe38c9133b4", | ||||
| 137 | + "evaluate_py": "fa6086aea49289d16cb8acae5cd0e8dfd2ecb057f89d75cb8fbf3f753065d8db" | ||||
| 138 | + }, | ||||
| 139 | + "F1-conf-04": { | ||||
| 140 | + "oracle_json": "c7280d2182d000f0220efae1c9559d91b63d7c1462579422e0ec20c605c67ea2", | ||||
| 141 | + "evaluate_py": "5dea91364ff7812220dc819283d5bb05fd8e3db943ce5e45bc8ab7b139b6466b" | ||||
| 142 | + }, | ||||
| 143 | + "F2-conf-01": { | ||||
| 144 | + "oracle_json": "d5628c5e913ac513d40d3de76007ffbfc44f8339cdecefe099263a0468e16d7a", | ||||
| 145 | + "evaluate_py": "e36ac3aa6a1bd89192c01c56c5ea5600e73ac34056489ebbefd3abb7045096f7" | ||||
| 146 | + }, | ||||
| 147 | + "F2-conf-02": { | ||||
| 148 | + "oracle_json": "8f18c929fa207eb352199d273647a7f28cf6680f01620e2bfae41131da465b00", | ||||
| 149 | + "evaluate_py": "d65e3565854ee1d9441f3363bf2cc765f3545a4ff0c73709709ea0f1832145dd" | ||||
| 150 | + }, | ||||
| 151 | + "F2-conf-03": { | ||||
| 152 | + "oracle_json": "0d666b81610441d8b771d554984e8e665d3a308abb89683967ed394d98e0c9e1", | ||||
| 153 | + "evaluate_py": "0ef5a00dbd8078646a381feb6d238e3fb90661a33c7a5a31b3e3c4bd0245dd1b" | ||||
| 154 | + }, | ||||
| 155 | + "F2-conf-04": { | ||||
| 156 | + "oracle_json": "2f7c8069533800ad85757f4aebbc92e77fb2dc5abf0551660ef43ba00efae589", | ||||
| 157 | + "evaluate_py": "bc4483e9b2d6b9e4a359fa008413873f4e3d6ff44a7512163e67d9d72e9619e3" | ||||
| 158 | + }, | ||||
| 159 | + "F3-conf-01": { | ||||
| 160 | + "oracle_json": "4bdc81b9911fd8af6b7eb61e560a8aed267a9e03218b07adc8afa72728fabe45", | ||||
| 161 | + "evaluate_py": "ae423d4660124829adb1a66a0ec68497578642b391012b66211a9a49fc76ba01" | ||||
| 162 | + }, | ||||
| 163 | + "F3-conf-02": { | ||||
| 164 | + "oracle_json": "eedec7d03dd243b4193827a51afdac386dae16fd206fe289174b0f9fb316ba82", | ||||
| 165 | + "evaluate_py": "c8b2d4d536137a142eb26d624bc7213f93edd43e70ca8cd5353e6a5ee31e0590" | ||||
| 166 | + }, | ||||
| 167 | + "F3-conf-03": { | ||||
| 168 | + "oracle_json": "273d2af8cf5c5aa4208c6c027b99b3bb991a50206b54c5d3c1a212dd610f3935", | ||||
| 169 | + "evaluate_py": "7daaedcbe4122426709899f1248e41add340fa7f7567064fa0b469784ba873da" | ||||
| 170 | + }, | ||||
| 171 | + "F3-conf-04": { | ||||
| 172 | + "oracle_json": "e4d1358d226ad618b91a80bf40bd5faca84e88da6cdfcca828d5c14cf5fc5e42", | ||||
| 173 | + "evaluate_py": "d451b45d0632cc775b8a7da767d1ef4865239f3ee4d0f5a8334746246cc48b3a" | ||||
| 174 | + }, | ||||
| 175 | + "F4-conf-01": { | ||||
| 176 | + "oracle_json": "6b1f40035072ca62d8892c68a75946607da2e2816432ecba6604ca0f0bef38e0", | ||||
| 177 | + "evaluate_py": "a2425e530cc89d68c8df54e41c60d569bb5c4ac4ce1ed0335eb76ca95be03d23" | ||||
| 178 | + }, | ||||
| 179 | + "F4-conf-02": { | ||||
| 180 | + "oracle_json": "89c2cd2985479ab67db1b9926dbaf37cc5f5a999132d26128e8bcec9aebbac76", | ||||
| 181 | + "evaluate_py": "a2425e530cc89d68c8df54e41c60d569bb5c4ac4ce1ed0335eb76ca95be03d23" | ||||
| 182 | + }, | ||||
| 183 | + "F4-conf-03": { | ||||
| 184 | + "oracle_json": "23998878a286573a4841e0b223c9efb51cf0be6bfb90da9d7fe40bc4f9200bb2", | ||||
| 185 | + "evaluate_py": "a2425e530cc89d68c8df54e41c60d569bb5c4ac4ce1ed0335eb76ca95be03d23" | ||||
| 186 | + }, | ||||
| 187 | + "F4-conf-04": { | ||||
| 188 | + "oracle_json": "6eb80dbbab7f408a8db45bef65c2a569fc1523d30f1976ec112f63c5bd621099", | ||||
| 189 | + "evaluate_py": "a2425e530cc89d68c8df54e41c60d569bb5c4ac4ce1ed0335eb76ca95be03d23" | ||||
| 190 | + }, | ||||
| 191 | + "F5-conf-01": { | ||||
| 192 | + "oracle_json": "b768e42a3731b47f1476771c4ba4e573b3e87f621f2ab4861067310c42cbeeed", | ||||
| 193 | + "evaluate_py": "63872700a684ede2b4178490b819738467b8bc08dc31f00b3d3d1be0b1cbe5c3" | ||||
| 194 | + }, | ||||
| 195 | + "F5-conf-02": { | ||||
| 196 | + "oracle_json": "f3f3f9d6197efad30f33fd30b7fd2af232f339da9f1fd4b5b538f298ec4214a5", | ||||
| 197 | + "evaluate_py": "e6d07df53bd6f319ff9ee731034d995862cbdc6835e02b40d8eed37146b942b8" | ||||
| 198 | + }, | ||||
| 199 | + "F5-conf-03": { | ||||
| 200 | + "oracle_json": "25dbf6ca4d4cc985ce472b7a50beb84157a1a03942bbcc8dc7efe5503db53319", | ||||
| 201 | + "evaluate_py": "0aec16a6bf2a076f73409766098ddd066dc7c305e9a250d128e1396885dc1532" | ||||
| 202 | + }, | ||||
| 203 | + "F5-conf-04": { | ||||
| 204 | + "oracle_json": "189c9db11884cf8611f6b8c28bbdcf5b2b97bdc212affe47fd3a5ac02dff3cb1", | ||||
| 205 | + "evaluate_py": "c5a9a24781c08a945e16eb8d669c3390ce03f2fee74850091917d38f32f0115b" | ||||
| 206 | + }, | ||||
| 207 | + "F6-conf-01": { | ||||
| 208 | + "oracle_json": "c8e71dcfdde7f49c5a79be5ada24f89bf758aefd8b4e29181b2be6175f01fef1", | ||||
| 209 | + "evaluate_py": "42f855b1eb4b9b15c799cc8b6f2ff6bc37d69fff511d3c02dce78153791daa43" | ||||
| 210 | + }, | ||||
| 211 | + "F6-conf-02": { | ||||
| 212 | + "oracle_json": "f7f2d808c8764007bf17157960edac7b5811a80cac28d6e6cf3aca547e81b162", | ||||
| 213 | + "evaluate_py": "2d13c09b9c3f71bd61a5509cc4f6ec262fc0d851634dbb7659027b2266d20365" | ||||
| 214 | + }, | ||||
| 215 | + "F6-conf-03": { | ||||
| 216 | + "oracle_json": "e605f4646a87a44aa733c28992c93adb555b9c9bed004be061cc3bef0506accf", | ||||
| 217 | + "evaluate_py": "e4f02908251872ba2e6a08b35f13f7c4887ff09b94a4351f6968ef60e21d3e07" | ||||
| 218 | + }, | ||||
| 219 | + "F6-conf-04": { | ||||
| 220 | + "oracle_json": "0c9223f343f191437ea26a918a2a77e21882e6c16a0a86b30fd200135eedc7b7", | ||||
| 221 | + "evaluate_py": "3fb94dc5e6e4109a51c2e9e472e98340204a202be63a0f6f7d803d4f9814236d" | ||||
| 222 | + } | ||||
| 223 | + }, | ||||
| 224 | + "protocol": { | ||||
| 225 | + "protocol/arms/arm_A.md": "7c4936f2e8a8125a7d6010476b4da8c4e4b8479d5f2ee4d505870a0744c35988", | ||||
| 226 | + "protocol/arms/arm_A_long.md": "f86279ae3194640c213873260108d6d43e0f31ed522c5db5370977e62572b793", | ||||
| 227 | + "protocol/arms/arm_B.md": "c2c8f685034a0915efa5f49dc682318def68bd00d2186d64094bd22b5be37322", | ||||
| 228 | + "protocol/arms/arm_C.md": "a779c9f93c477d08fb6b0138571fb50c913289955929614db80e1231c082e9c0", | ||||
| 229 | + "protocol/arms/arm_D.md": "3d751928296c191bbf40500a9291436e6578bf2f6a908a006bd6416d4fb5399c", | ||||
| 230 | + "protocol/arms/arm_L.md": "f86279ae3194640c213873260108d6d43e0f31ed522c5db5370977e62572b793", | ||||
| 231 | + "protocol/arms/common_preamble.md": "935b2032e111f43198a46a5f516d5d871f47884d1da8a966e1dbf5a7bc8f9436", | ||||
| 232 | + "protocol/arms/gate_feedback_prompt.md": "ecf6ce7a9bf2e142bb92eeff30c0d7f1e14dd3dfd5171eacbc44205759be15a2", | ||||
| 233 | + "protocol/arms/resume_after_interruption_prompt.md": "6dc1ee35df4d764358a3e64e5dec8bf4442cd2a261c09a44e1128f676326f0b4", | ||||
| 234 | + "protocol/arms/review_feedback_prompt.md": "ea08bbc6b85e8c9454a851e6c8b91856cc1c0e045ebdfd9883faac608028205b", | ||||
| 235 | + "protocol/arms/reviewer.md": "83828203e4b95c16d2ac1e764aabc7c7f3e2840eda7c7c93d5e4edc51f335da5", | ||||
| 236 | + "protocol/arms/task_prompt.md": "a456de794c9f05dacc38a7b8a29fe87ded23647c7576cdc0e47e66c3be79f51d", | ||||
| 237 | + "protocol/preregistration.md": "7562e596e4ebeb5e372b3447a69ea89c9f61b2d47bce840820fe96b02cd8d260", | ||||
| 238 | + "src/pgar/runtime_gates.py": "1760d13e9f5bf9cfba92a8c6cd4ceb470cad06fec7759eab21463f5ce44b92f0", | ||||
| 239 | + "src/pgar/launcher.py": "e27532fc73cf17128c36b90a4f2d1156947d6973e0d73570be13e585cc3f2e10", | ||||
| 240 | + "src/pgar/scheduler.py": "8e3d05b626686ff8567af7db3f6b5078c1a2b86a9b12a8de5d4db1ea095d4760", | ||||
| 241 | + "src/pgar/evaluation.py": "100d3fcdb0bafdc822431179c28aa8ee97f88cea474a1bdf256647f566fd0faf" | ||||
| 242 | + }, | ||||
| 243 | + "configs": { | ||||
| 244 | + "configs/budgets.yaml": "412bb281571e7138b248d5f6de36577b44c64a1604f48ba62b04d1e5296e2ac4", | ||||
| 245 | + "configs/capture.yaml": "64ed330a21d714ee71fab2c88c28a7459a792f362a03c337709a8b6e2c4a8513", | ||||
| 246 | + "configs/confirmatory.yaml": "0851d028436d127f50e81a7b3690a4c7ca8b3def678965e0e089c386f6fa88aa", | ||||
| 247 | + "configs/development.yaml": "4087799aa303a197b7703e5b1f8a99f4da5aaaaf481942a375561bef1e7f80ca", | ||||
| 248 | + "configs/pilot.yaml": "43b8510075ae798cee745b792f47d499256b51e3583639689d49004625299904", | ||||
| 249 | + "configs/transfer.yaml": "da394c10c0114be49d30877d8e91fb533c223a9ef4a51aa12dd10232eb653ac9" | ||||
| 250 | + }, | ||||
| 251 | + "analysis": { | ||||
| 252 | + "analysis/build_analysis_dataset.py": "b790f14e0538ea1d2ae36db4f1486d0eadb6f6a7fee029a7230d787cc60c33d7", | ||||
| 253 | + "analysis/common.py": "a93ebd683f7b437df540d52f593f2208552c589beb9de9103f06771072b0bbb7", | ||||
| 254 | + "analysis/figures.py": "d234a24f388fda23416f709d6d49fcaa5ff4833409b8ff324d79a7065a10acde", | ||||
| 255 | + "analysis/paper_numbers.py": "3b8d70cecdc21e083df61c813961343e6faadc90afe8ce9ceab0b33dfdd0c773", | ||||
| 256 | + "analysis/pilot_report.py": "8ed65668194960835dd30e9bf2503df311b0281de6e04990070c3e1042c39ffb", | ||||
| 257 | + "analysis/primary_analysis.py": "bca632aa9bf167940d22cbe417050dbd3002a8ab2607edb49c50589a084834e0", | ||||
| 258 | + "analysis/robustness.py": "8f432bd85dbadfcf478aac5d178af0b5ca4734e3f8d9147ad4893a778ed3d7c1", | ||||
| 259 | + "analysis/tables.py": "3a949e8a8c6ba8f1983574c30d11b3182e45872aa1ec77ab85ccf215c0156766", | ||||
| 260 | + "analysis/task_tables.py": "c11d2955ea2a94038f85f5cf1b1625ee36849936676077881f87c6c390beaefd" | ||||
| 261 | + }, | ||||
| 262 | + "code_commit": "90bf126799e046545d1bf5210fce451071dce757" | ||||
| 263 | +} | ||||
| \ No newline at end of file | |||||
added
runs/confirmatory/assignment_seed301.json
+10946 −0
Large diff (10,947 lines) is not loaded by default to keep the page fast.Load diff
modified
runs/state.sqlite
+0 −0
Binary file not shown.
added
runs/state.sqlite-shm
+0 −0
Binary file not shown.
added
runs/state.sqlite-wal
+0 −0
Large diff (0 lines) is not loaded by default to keep the page fast.Load diff
modified
tasks/registry.csv
+24 −24
Large diff (57 lines) is not loaded by default to keep the page fast.Load diff