agent-harness / configs /study5 /E16_selection.json
cuber12's picture
Publish agent harness research code and paper artifacts
d61821a verified
Raw
History Blame Contribute Delete
6.81 kB
{
"decisions": [
{
"harness_id": "H000",
"role": "mandatory_exact_baseline"
},
{
"harness_id": "H007",
"role": "highest_resolution"
},
{
"harness_id": "H002",
"role": "lowest_tokens_within_5pp_resolution"
},
{
"harness_id": "H014",
"role": "highest_accepted_edit"
},
{
"harness_id": "H008",
"role": "best_prespecified_robustness_candidate"
},
{
"harness_id": "H001",
"role": "pareto_frontier_fill"
}
],
"e16_manifest_sha256": "ab444d5449b4b0cafce87c1afeb0036dd546496c824467ee40e7a2c79fbacb88",
"e16_planned_cells": 306,
"excluded_screening_experiment": {
"experiment_id": "E14",
"reason": "retrieval harnesses are repeated across action interfaces and are reserved for retrieval-by-action inference"
},
"metrics": {
"H000": {
"accepted_edit_rate": 0.4666666666666667,
"applicable_patch_rate": 0.46111111111111114,
"cells": 180,
"mean_total_tokens": 72646.88333333333,
"mean_wall_seconds": 41.005881560401114,
"resolved_rate": 0.1
},
"H001": {
"accepted_edit_rate": 0.42777777777777776,
"applicable_patch_rate": 0.4111111111111111,
"cells": 180,
"mean_total_tokens": 72252.48333333334,
"mean_wall_seconds": 40.23372465622508,
"resolved_rate": 0.07777777777777778
},
"H002": {
"accepted_edit_rate": 0.3888888888888889,
"applicable_patch_rate": 0.37777777777777777,
"cells": 180,
"mean_total_tokens": 61987.822222222225,
"mean_wall_seconds": 40.5383593710411,
"resolved_rate": 0.08888888888888889
},
"H003": {
"accepted_edit_rate": 0.4222222222222222,
"applicable_patch_rate": 0.4166666666666667,
"cells": 180,
"mean_total_tokens": 74789.28888888888,
"mean_wall_seconds": 47.87801928565734,
"resolved_rate": 0.12222222222222222
},
"H004": {
"accepted_edit_rate": 0.46111111111111114,
"applicable_patch_rate": 0.45,
"cells": 180,
"mean_total_tokens": 68463.7,
"mean_wall_seconds": 47.970295593952244,
"resolved_rate": 0.1111111111111111
},
"H005": {
"accepted_edit_rate": 0.4111111111111111,
"applicable_patch_rate": 0.40555555555555556,
"cells": 180,
"mean_total_tokens": 74441.46666666666,
"mean_wall_seconds": 43.979250848839165,
"resolved_rate": 0.09444444444444444
},
"H006": {
"accepted_edit_rate": 0.45555555555555555,
"applicable_patch_rate": 0.4388888888888889,
"cells": 180,
"mean_total_tokens": 74994.5,
"mean_wall_seconds": 48.37345648450104,
"resolved_rate": 0.10555555555555556
},
"H007": {
"accepted_edit_rate": 0.43333333333333335,
"applicable_patch_rate": 0.42777777777777776,
"cells": 180,
"mean_total_tokens": 72679.36666666667,
"mean_wall_seconds": 47.19653170536427,
"resolved_rate": 0.12222222222222222
},
"H008": {
"accepted_edit_rate": 0.38333333333333336,
"applicable_patch_rate": 0.38333333333333336,
"cells": 60,
"mean_total_tokens": 66450.58333333333,
"mean_wall_seconds": 36.10224259162787,
"resolved_rate": 0.0
},
"H009": {
"accepted_edit_rate": 0.38333333333333336,
"applicable_patch_rate": 0.38333333333333336,
"cells": 60,
"mean_total_tokens": 63343.78333333333,
"mean_wall_seconds": 33.317234545220465,
"resolved_rate": 0.0
},
"H010": {
"accepted_edit_rate": 0.4166666666666667,
"applicable_patch_rate": 0.36666666666666664,
"cells": 60,
"mean_total_tokens": 87210.51666666666,
"mean_wall_seconds": 81.18024050413321,
"resolved_rate": 0.0
},
"H011": {
"accepted_edit_rate": 0.26666666666666666,
"applicable_patch_rate": 0.25,
"cells": 60,
"mean_total_tokens": 65144.8,
"mean_wall_seconds": 28.2178086270617,
"resolved_rate": 0.0
},
"H012": {
"accepted_edit_rate": 0.4166666666666667,
"applicable_patch_rate": 0.4,
"cells": 60,
"mean_total_tokens": 103753.21666666666,
"mean_wall_seconds": 48.84829152571425,
"resolved_rate": 0.0
},
"H013": {
"accepted_edit_rate": 0.36666666666666664,
"applicable_patch_rate": 0.36666666666666664,
"cells": 60,
"mean_total_tokens": 77243.36666666667,
"mean_wall_seconds": 38.010123237486305,
"resolved_rate": 0.0
},
"H014": {
"accepted_edit_rate": 0.4666666666666667,
"applicable_patch_rate": 0.45,
"cells": 60,
"mean_total_tokens": 145122.58333333334,
"mean_wall_seconds": 65.632639020142,
"resolved_rate": 0.0
},
"H015": {
"accepted_edit_rate": 0.35,
"applicable_patch_rate": 0.35,
"cells": 60,
"mean_total_tokens": 63847.38333333333,
"mean_wall_seconds": 35.46575253612439,
"resolved_rate": 0.0
}
},
"pareto_frontier": [
"H000",
"H001",
"H002",
"H004",
"H007",
"H009",
"H011"
],
"primary_screening_panel": {
"H000": "E13",
"H001": "E13",
"H002": "E13",
"H003": "E13",
"H004": "E13",
"H005": "E13",
"H006": "E13",
"H007": "E13",
"H008": "E15",
"H009": "E15",
"H010": "E15",
"H011": "E15",
"H012": "E15",
"H013": "E15",
"H014": "E15",
"H015": "E15"
},
"robustness_candidates": [
"H008",
"H010"
],
"schema_version": 1,
"screening_experiments": [
"E13",
"E15"
],
"screening_revision": "7f4de67853deab34aca5a9ceaaf7e9f85b901088",
"selected_harnesses": [
"H000",
"H001",
"H002",
"H007",
"H008",
"H014"
],
"selection_is_outcome_aware_to_screening": true,
"selection_is_outcome_blind_to_e16": true,
"selection_sha256": "a6604699ba76e9eb95d31d5987363af0acd66d7ee0ed7d20e2eb64a7806b2e02",
"source_final_reports": {
"E13": {
"accepted_edit_count": 624,
"applicable_patch_count": 610,
"path": "results/reports/E13_7f4de67853de_1784618186.json",
"resolved_count": 148,
"run_count": 1440,
"sha256": "98cc58a125b5aa5c69e90b46134d9e01a93bd37d7d4f5d2160e0921746fb29a8"
},
"E15": {
"accepted_edit_count": 206,
"applicable_patch_count": 199,
"path": "results/reports/E15_7f4de67853de_1784664649.json",
"resolved_count": 0,
"run_count": 540,
"sha256": "842cb293fd4269de8fafdb97b343d3baf78c789a54b94e8b3136d75216e83dab"
}
},
"source_raw_sha256": {
"E13": "723a554fdb2596f0f8ece3d86f3f9b4333d2589bdae6e638b40798cb3b4c775e",
"E15": "30d40d4d371a8dac7554648f0066d85660d5486b3db8de686f7b66911d9f6e28"
},
"study": "Study 5 end-to-end harness behavior"
}