1{2 "frozen_at": "2026-10-28T01:05:40.717943+00:00",3 "package_version": "0.1.0",4 "git_commit": null,5 "python": "3.14.4",6 "role_prompt_sha256": {7 "appraisal_method": "4ab6a4b56e275e66ceab45a7ec123dbb971d67aa5a6408f9398e58c1cdc36e2a",8 "comparables": "831b81649f59fa7ebeeff549643dfc73c0738cf7e2cf8116321628a153b77ee6",9 "hedonic": "9063d328055c29d64ce605a64de09719e99fc7b9d0332c4b7ab289c10fa8a0d5",10 "market_economist": "623f2000f584463efd0b52f0c18f61b777ffd07a110008ec7033a48608de659b",11 "building_condition": "d4c83732b1f6dc260f6e37dd7fe7b86ea17b3a1fa1fb9636204247ddda7c1f91",12 "geospatial_risk": "83d7221633a884c3f04ad5ce6b2917bf274c206da50c693062f71469fc4af486",13 "underwriting": "d25a69fc8ead39ca2e35b9e938c0420b23fe8581936862b286d5149b9e7050a2",14 "buyer_wtp": "facb211deeccae1ce37ae61efb2aa5244be533c114d794290dc02d38b5d74663",15 "seller_strategy": "3917bd2ffd8e55431d35689768f9070398c5fbc21c45860eea6bcdcfce7c2053",16 "rental_usercost": "e745f4c42bb39244347b553b08b1d732e42bfdbf533ccbf385cf92664305f6af",17 "data_auditor": "db248eb5f189de2b0c2605c8330f05ddae1f86d8bc72fccdab449bce17517d5c",18 "skeptic": "2f789b8366c5d29c8d522f036b970421304d772c71a85857c29b9d4214dad629",19 "construction_cost": "662fa67c232fac1604d0205cd2906b5d1983b2f8935bd39ae34823180b46417d",20 "zoning": "b67c5b6be3f435d090a12d2dd79999a1c0a61ff6d165ec35983703056f4a3530",21 "macro_rates": "07c1d1f7dca94f66fed09ad6516e786c1cfaed025c927577bf045465f2db1438",22 "energy": "b1577fe8ad0f7724111c5d879f74e36a9741b65b8ba245fe6b5f7145552ed1f5",23 "climate_risk": "b18c0d82f1056a025ef8f3fe0f060414784d64e1c7cbb1cbccc414fbf8748d51",24 "regional_expert": "a422302dc0ab198d040b1bddb177e3dba2d486252c1583b2aae1836ce40a888e",25 "uncertainty_auditor": "b4f063d44da5c6c7c98ebf235566c15b16f9e39738d5e26bdd95fe5f92fd4fb2",26 "fairness_auditor": "70596b6f995f3533b8e56db09563c3b19fa198590188f6097785e48ec3f28b90",27 "generalist": "ade12bb690d31dbead8b869e7c0f86fc290bbdcf930a7736696953a90079ee82",28 "placebo_1": "2e639018993a7f909df72239c0655df5ff2435bea755b6e727684ae316551e02",29 "placebo_2": "3b8dc64ee9d7510c3e38d66eaa2fd2dc35b0fa98b9813b9801043d35cdaeeb53",30 "placebo_3": "e53e5b360a75a3d76de904e43424e572a555b4b60482176092268de9c64dbc84",31 "placebo_4": "c7ad48395054051df732dfd2acfb742f95bfc1b088abb8ca16e8ace6fd10919d",32 "placebo_5": "60f3ff16d47dee84d94e4faa57b8708a453e3ce054fa752313de5ebf88075f19",33 "placebo_6": "25278ebbadf591ea1e1e56298dcd2e78dadc9d72b5499440402f83b7a3edfb72",34 "placebo_7": "6ca049c56da54247d50342866fc38f5cd4cf7a5ffd441f8dcf9c28bcd374002e",35 "placebo_8": "79f63b31af873c44e4cdec4dd5ac3827f3894992b587e4f26cd5b96c31319011",36 "placebo_9": "f615020da100bf33b8f3ab823cf7cbe5a832cbb723224ada2368d734f5d9462e",37 "placebo_10": "924b26c724c0e3584d2ad4dd86c7d813d914c8dff64a560eadc763253e0976b3",38 "placebo_11": "fcb29d21820098bd3045855444f5c586680dc5fb96eb663f6486e4baacc55291",39 "placebo_12": "0866555d436b3cc8768ad7f3fec71b4943add5529b8f700ec502e16aeddebea8"40 },41 "role_prompt_hash16": {42 "appraisal_method": "4ab6a4b56e275e66",43 "comparables": "831b81649f59fa7e",44 "hedonic": "9063d328055c29d6",45 "market_economist": "623f2000f584463e",46 "building_condition": "d4c83732b1f6dc26",47 "geospatial_risk": "83d7221633a884c3",48 "underwriting": "d25a69fc8ead39ca",49 "buyer_wtp": "facb211deeccae1c",50 "seller_strategy": "3917bd2ffd8e5543",51 "rental_usercost": "e745f4c42bb39244",52 "data_auditor": "db248eb5f189de2b",53 "skeptic": "2f789b8366c5d29c",54 "construction_cost": "662fa67c232fac16",55 "zoning": "b67c5b6be3f435d0",56 "macro_rates": "07c1d1f7dca94f66",57 "energy": "b1577fe8ad0f7724",58 "climate_risk": "b18c0d82f1056a02",59 "regional_expert": "a422302dc0ab198d",60 "uncertainty_auditor": "b4f063d44da5c6c7",61 "fairness_auditor": "70596b6f995f3533",62 "generalist": "ade12bb690d31dbe"63 },64 "optional_roles": [65 "construction_cost",66 "zoning",67 "macro_rates",68 "energy",69 "climate_risk",70 "regional_expert",71 "uncertainty_auditor",72 "fairness_auditor"73 ],74 "configs_sha256": {75 "budgets.yaml": "d6d368fd34b923cead2174f9f233f6d0c5ed8dce8885f12fce02372bb43dfc73",76 "data.yaml": "c20f448c79e39726f9c5faf5be9da72ee4d963213a7dcd807f6b48638d4f2565",77 "models.yaml": "80e216f864e413ceafa547430ebf885b27df7efce2c993d00578c2d441fa8c9c",78 "treatments.yaml": "e1f71e68e67f7701093396370507542c622deac316878b2a356937c0dfc6ad0f"79 },80 "code_sha256": {81 "agents/__init__.py": "f08361539ddc903d3082f282b7beca392daafa081d9d29b735e7ffacaff76e92",82 "agents/packets.py": "7cc67995151343bbcf3bc516c64ada1f7291aca7ac3278651c5f56175ad08f09",83 "agents/roles.py": "7ea4c36eba414138a09039291aad98595de5a0c10c32833dffc0d360629f8db9",84 "agents/runner.py": "93e97a2ce6d26d6a27b48a5dc41ce0beb3c17ebbfe849684d99b5e37e4cf01b6",85 "debate/__init__.py": "e2abecdc8cabcef2e6aed8d0e53d35064b1e16d1e494f1d8f6393fa6d197947e",86 "debate/aggregation.py": "082ef25310267f3b2f262cec30c967a116d11b3f5a284654f96cd83b7c8d09d4",87 "debate/committees.py": "3ab3616d6839999d0f48ed9b4842dc8694e9f75cbf911a02d660dd885c6b5603",88 "debate/protocols.py": "1c7a3160c62cab1728e329701a5d1d4c028c8689b8dff977cb4b3ea1732e8cfe",89 "debate/stopping.py": "9d1202180ebb7713153b36548cff48f0e601520780ea47603207c31ad910ea4c",90 "schemas.py": "b0170725f58da4edd21a11dfb8905f37422d48cf8d6fa98901b9fb46f84c7e67",91 "providers/__init__.py": "7dac58387e3ea2f79e9d17263a3f1c913c66d7a35812d1d0050e101c24619108",92 "providers/anthropic_provider.py": "d5b456b942e593ea87e1f542fdf99eb1bbc923a0bdad616a007bb0ada34c996e",93 "providers/base.py": "8b1f06b7d33aad7e246b5612ca42d45909de0483e0b29222660e28c204809818",94 "providers/ledger.py": "c9c231b73bb358a4a43035624c825a6b708271e313953c2ee93b01f69afdbd2c",95 "providers/mock.py": "cffe47ff9a9ba243b1415887b593546b44f080264b98c3814290983870144337",96 "providers/openai_provider.py": "67171aafbdc20f6eada9a3cb1d1f9a3a1f816ccde27e65fa158e2ff52de0c9c4"97 },98 "model_ids": {99 "cost_matched_pairs": {100 "mid": {101 "anthropic": "claude-sonnet-5-5",102 "openai": "gpt-6.1-sol"103 },104 "small": {105 "anthropic": "claude-haiku-5-5",106 "openai": "gpt-6-luna"107 }108 },109 "default_tier": "mid",110 "request_defaults": {111 "anthropic": {112 "max_tokens": 4096,113 "effort": "medium",114 "thinking": "adaptive"115 },116 "openai": {117 "max_output_tokens": 4096,118 "reasoning_effort": "medium"119 },120 "timeout_s": 180,121 "max_retries": 4,122 "concurrency": 10123 }124 },125 "treatments": {126 "arms": {127 "A_hedonic": {128 "family": "baseline",129 "model": "hedonic_ridge",130 "note": "log-price OLS/ridge with municipality x quarter and grid fixed effects"131 },132 "B_hierarchical": {133 "family": "baseline",134 "model": "hierarchical_hedonic",135 "note": "mixed-effects hedonic with municipality and grid random intercepts; repeat-sales adjustment when the parcel has a prior sale"136 },137 "C_gbm": {138 "family": "baseline",139 "model": "lightgbm",140 "note": "LightGBM tuned on validation only; CatBoost/XGBoost reported as robustness"141 },142 "D_avm": {143 "family": "baseline",144 "model": "avm",145 "note": "GBM + comparable-based features + spatial-temporal lags + conformal / quantile intervals"146 },147 "E_single_llm": {148 "family": "llm",149 "roles": [150 "generalist"151 ],152 "rounds": 0153 },154 "F_independent": {155 "family": "llm",156 "roles": "default12",157 "rounds": 0,158 "aggregation": "all"159 },160 "G_debate": {161 "family": "llm",162 "roles": "default12",163 "rounds": [164 1,165 2,166 4167 ],168 "protocol": "structured_debate"169 },170 "H_debate_skeptic": {171 "family": "llm",172 "roles": "default12",173 "rounds": 2,174 "protocol": "structured_debate",175 "blinded_skeptic": true,176 "evidence_verification": true177 },178 "I_placebo": {179 "family": "llm",180 "roles": "shuffled12",181 "rounds": 2,182 "protocol": "structured_debate",183 "note": "role prompts randomly permuted / generic placeholder roles"184 },185 "G_fullview": {186 "family": "llm",187 "roles": "default12_fullview",188 "rounds": 2,189 "protocol": "structured_debate",190 "note": "secondary contrast: default prompts, full packet for every role (value of the evidence partition)"191 },192 "J_cost_matched": {193 "family": "llm",194 "roles": [195 "generalist"196 ],197 "repeats": 12,198 "rounds": 0,199 "note": "single agent sampled k times to match the token budget of F/G; plus self-refine chain"200 }201 },202 "aggregation_rules": [203 "mean",204 "median",205 "trimmed_mean_20",206 "inverse_variance",207 "stacking",208 "robust_bayes"209 ],210 "rounds_grid": [211 0,212 1,213 2,214 4215 ],216 "providers": {217 "committees": [218 "all_anthropic",219 "all_openai",220 "mixed_random"221 ],222 "tier": "mid"223 }224 },225 "freeze_digest": "f0e9d6a94b0387f92645346af2c8c63200c8b9f83c9a85095f507cffa2e85e1c"226}