{
  "protocol": {
    "episodes": 120,
    "memory_size": 24,
    "seed": 42,
    "task_source": "deterministic public-style mini-suites; not original API/game benchmarks"
  },
  "results": {
    "self-refine": {
      "benchmark": "planbench-mini",
      "joint_success": 1.0,
      "average_cost": 2.0,
      "refinements": 120,
      "critic_rounds": 120,
      "reasoning_steps": 240
    },
    "rewoo": {
      "benchmark": "planbench-mini",
      "joint_success": 1.0,
      "average_cost": 3.5,
      "plans_created": 120,
      "worker_calls": 360,
      "actions": 360
    },
    "autogen": {
      "benchmark": "planbench-mini",
      "joint_success": 1.0,
      "average_cost": 3.0,
      "plans_created": 120,
      "agent_messages": 360,
      "critic_rounds": 120
    },
    "pearl": {
      "benchmark": "planbench-mini",
      "joint_success": 1.0,
      "average_cost": 1.12,
      "plans_created": 12,
      "plan_explorations": 24,
      "policy_updates": 12,
      "reused_plans": 108
    },
    "tree-of-thoughts": {
      "benchmark": "planbench-mini",
      "joint_success": 1.0,
      "average_cost": 2.5,
      "tree_nodes_expanded": 1200,
      "backtracks": 480
    },
    "lats": {
      "benchmark": "planbench-mini",
      "joint_success": 1.0,
      "average_cost": 4.0,
      "search_rollouts": 480,
      "reflections": 360,
      "backtracks": 360
    },
    "toolformer": {
      "benchmark": "scalemcp-mini",
      "joint_success": 1.0,
      "average_cost": 3.0,
      "tool_call_candidates": 540,
      "tool_calls_accepted": 360
    },
    "react": {
      "benchmark": "scalemcp-mini",
      "joint_success": 1.0,
      "average_cost": 3.0,
      "reasoning_steps": 360,
      "actions": 360
    },
    "reflexion": {
      "benchmark": "planbench-mini",
      "joint_success": 0.9,
      "average_cost": 1.1,
      "reflections": 12,
      "reused_plans": 108
    },
    "voyager": {
      "benchmark": "planbench-mini",
      "joint_success": 1.0,
      "average_cost": 1.12,
      "skills_created": 12,
      "skills_reused": 108,
      "verification_retries": 12
    }
  },
  "schema_version": 2,
  "manifest_ref": "experiments:classic-agent-mini-suites-seed42",
  "evaluation_protocol": {
    "tier": "l1_mechanism",
    "seeds": [
      42
    ],
    "formal_comparison": false,
    "claim_policy": "single/few-seed smoke result; do not claim a stable improvement"
  },
  "provenance": {
    "artifact_path": "docs/experiments/classic-agent-mini-suites-seed42.json",
    "historical_migration": "historical-metrics-v2-2026-08-09",
    "original_code_commit": "not recorded",
    "dataset_fingerprint": "not recorded in historical artifact"
  }
}