{
  "schema_version": 1,
  "evidence_tier": "real_minihack_environment_smoke",
  "paper_result_reproduced": false,
  "controller": "seeded_uniform_random",
  "environment": "MiniHack-Room-5x5-v0",
  "seeds": [42, 43, 44],
  "modes": ["primitives", "skills", "mixed"],
  "budgets": {"max_decisions": 8, "primitive_budget": 64},
  "upstream_codehack_commit": "3026cdb7378726eace9efe8ab655f1aee616844f",
  "upstream_baselines_reference_only": "ba2c532bd505426913b82291adec455afd855df5",
  "upstream_nle_wheel_sha256": "819063e40b3c1bc97e946695242b7cfe0633d61049ac44cfde53d72447948513",
  "episodes": 9,
  "errors": 0,
  "primitive_steps_by_mode_and_seed": {
    "primitives": [1, 1, 1],
    "skills": [20, 20, 20],
    "mixed": [0, 0, 20]
  },
  "terminated_by_mode_and_seed": {
    "primitives": [false, false, false],
    "skills": [true, true, true],
    "mixed": [false, false, true]
  },
  "scope": "Action-space and real-environment integration only. Modes use different random action spaces; these rates are not a matched policy comparison, paper improvement, or RL/LLM result."
}