File size: 3,515 Bytes
e8ba6c7 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 | {
"schema_version": "savc.public.depth_study.v1",
"status": "COMPLETE_CRITERIA_NOT_MET",
"study": {
"selected_episode_count": 60,
"condition_count": 9,
"model_family_count": 2,
"repeat_count": 2,
"scheduled_trace_count": 2160,
"valid_trace_count": 2157,
"invalid_trace_count": 3,
"human_relabeling_performed": false
},
"execution_integrity": {
"evaluation_labels_exposed_to_generation": false,
"scientific_outputs_inspected_before_scoring_rules_fixed": false,
"role_isolation_violations": 0,
"all_parseability_requirements_met": true,
"initial_output_cap_tokens": 500,
"qualified_output_cap_tokens": 1000,
"output_cap_qualification_used_scientific_results": false,
"only_failed_slots_regenerated_after_quota_errors": true,
"scoring_api_calls": 0,
"trace_regeneration_during_scoring": false
},
"prespecified_criteria": {
"handoff_failure_exists": "FAIL",
"depth_response_exists": "FAIL",
"receipts_preserve_evidence_in_both_model_families": "FAIL",
"detachment_has_action_consequence_in_both_model_families": "FAIL",
"runtime_gate_reduces_harm": "NOT_ESTIMABLE"
},
"handoff_results": {
"openai_family": {
"unstructured_depth4_final_evidence_detachment_rate": 1.0,
"structured_depth4_final_evidence_detachment_rate": 0.291667,
"unstructured_minus_structured_risk_difference": 0.708333,
"paired_bootstrap_95_ci": [
0.608333,
0.8
]
},
"anthropic_family": {
"unstructured_depth4_final_evidence_detachment_rate": 1.0,
"structured_depth4_final_evidence_detachment_rate": 1.0,
"unstructured_minus_structured_risk_difference": 0.0,
"paired_bootstrap_95_ci": [
0.0,
0.0
],
"structured_responsibility_survival_by_hop": {
"1": 1.0,
"2": 0.0,
"3": 0.0,
"4": 0.0
}
}
},
"action_result": {
"evaluator_labeled_commit_eligible_traces": 648,
"observed_committed_actions": 17,
"useful_commits": 0,
"unsupported_commits": 17,
"root_positive_or_negative_status_count": 0,
"action_effect_interpretable": false
},
"input_audit": {
"evaluator_labeled_commit_eligible_episode_count": 18,
"empty_visible_evidence_count": 9,
"source_identifier_only_visible_evidence_count": 9,
"direct_result_payload_count": 0,
"claims_requesting_tool_use": 18,
"scientific_tool_interface_available": false,
"action_result_interpretable_from_provided_inputs": false
},
"result_classification": {
"primary_claim": "The depth and action study did not meet its prespecified success criteria across both model families.",
"supported_secondary_claim": "Structured receipts reduced depth-four evidence detachment in one model family but did not transfer to the second family.",
"failure_diagnosis": "Agents were not given result-bearing evidence in any commit-eligible episode, so the action result could not be interpreted.",
"biological_discovery_claim_supported": false,
"soft_semantic_verifier_validated": false
},
"release_boundary": {
"aggregate_only": true,
"raw_traces_in_output": false,
"row_level_scores_in_output": false,
"episode_ids_in_output": false,
"source_ids_in_output": false,
"hidden_labels_in_output": false,
"local_paths_in_output": false,
"credentials_in_output": false,
"biological_design_instructions_in_output": false
},
"study_id": "depth_study"
}
|