debatefloor / reports /http_rollout_eval.json
AniketAsla's picture
sync: mirror git d05fcb5 to Space
b4ac377 verified
{
"base_url": "https://aniketasla-debatefloor.hf.space",
"rows": [
{
"policy": "naive_high_no_investigation",
"task_id": "clean_claim",
"seed": 42,
"steps": 1,
"done": true,
"reward": 0.7998,
"final_decision": "approve_claim",
"agent_confidence": "HIGH",
"calibration_score": 1.0,
"decision_accuracy": 1.0,
"fraud_detection_score": 1.0,
"evidence_quality_score": 1.0
},
{
"policy": "naive_high_no_investigation",
"task_id": "contradictory_claim",
"seed": 42,
"steps": 1,
"done": true,
"reward": 0.0,
"final_decision": "approve_claim",
"agent_confidence": "HIGH",
"calibration_score": -0.8,
"decision_accuracy": 0.0,
"fraud_detection_score": 0.0,
"evidence_quality_score": 0.0
},
{
"policy": "naive_high_no_investigation",
"task_id": "distribution_shift_claim",
"seed": 42,
"steps": 1,
"done": true,
"reward": 0.0,
"final_decision": "approve_claim",
"agent_confidence": "HIGH",
"calibration_score": -0.8,
"decision_accuracy": 0.0,
"fraud_detection_score": 0.0,
"evidence_quality_score": 0.0
},
{
"policy": "calibrated_scripted_investigator",
"task_id": "clean_claim",
"seed": 42,
"steps": 4,
"done": true,
"reward": 0.7623,
"final_decision": "approve_claim",
"agent_confidence": "HIGH",
"calibration_score": 1.0,
"decision_accuracy": 1.0,
"fraud_detection_score": 1.0,
"evidence_quality_score": 1.0
},
{
"policy": "calibrated_scripted_investigator",
"task_id": "contradictory_claim",
"seed": 42,
"steps": 7,
"done": true,
"reward": 0.5468,
"final_decision": "deny_claim",
"agent_confidence": "MED",
"calibration_score": 0.6,
"decision_accuracy": 1.0,
"fraud_detection_score": 0.75,
"evidence_quality_score": 0.0
},
{
"policy": "calibrated_scripted_investigator",
"task_id": "distribution_shift_claim",
"seed": 42,
"steps": 8,
"done": true,
"reward": 0.3522,
"final_decision": "escalate_to_human",
"agent_confidence": "LOW",
"calibration_score": 0.1,
"decision_accuracy": 1.0,
"fraud_detection_score": 0.0,
"evidence_quality_score": 0.0
}
]
}