{
  "schema_version": 2,
  "manifest_ref": "agent-research:user-proxy-bench",
  "method": "user-proxy-bench",
  "dataset": "deterministic public mechanism mini-suite",
  "seeds": [
    42,
    43,
    44
  ],
  "diagnostic_only": true,
  "metrics": {
    "user_fidelity_score": {
      "mean": 0.6666666666666666,
      "std": 0.23570226039551584
    },
    "episodes": {
      "mean": 2.0,
      "std": 0.0
    }
  },
  "seed_results": [
    {
      "seed": 42,
      "user_fidelity_score": 0.5,
      "episodes": 2,
      "agent_reward_used": false
    },
    {
      "seed": 43,
      "user_fidelity_score": 1.0,
      "episodes": 2,
      "agent_reward_used": false
    },
    {
      "seed": 44,
      "user_fidelity_score": 0.5,
      "episodes": 2,
      "agent_reward_used": false
    }
  ],
  "evaluation_protocol": {
    "tier": "l1_mechanism",
    "seeds": [
      42,
      43,
      44
    ],
    "formal_comparison": false,
    "diagnostic_only": true,
    "claim_policy": "executable mechanism and invariant checks only"
  },
  "provenance": {
    "artifact_path": "docs/agent-research/2609.38043-userproxybench/metrics/mechanism-seeds42-44.json",
    "dataset_fingerprint": "deterministic public mechanism suite sep30-followup-v1",
    "original_code_commit": "working tree"
  }
}
