{
  "run_dir": "/data4/toby/Branching-Search/runs/agents/e1_physics_dev",
  "cell": {
    "run": "e1_physics_dev",
    "kind": "physics",
    "policy": "uniform",
    "N": 8,
    "M": 8,
    "K": 4.0,
    "V": 32.0,
    "T": 4,
    "rounds": 3,
    "b": 2,
    "m": 1,
    "uniform_families": true,
    "tcas_evidence": null
  },
  "rounds": 3,
  "degrees": 2,
  "degrees_growing": [
    0,
    2
  ],
  "degrees_decaying": [],
  "separation_is_monotone": false,
  "endpoint_crossing": {
    "status": "no_crossing_all_growing",
    "d_c": null
  },
  "lineage_crossing": {
    "d_c": null,
    "status": "no_crossing_all_positive",
    "ci": [
      null,
      null
    ]
  },
  "figures": [
    "/data4/toby/Branching-Search/runs/agents/e1_physics_dev/figures/order_parameter_A_t.png",
    "/data4/toby/Branching-Search/runs/agents/e1_physics_dev/figures/order_parameter_A_t.pdf",
    "/data4/toby/Branching-Search/runs/agents/e1_physics_dev/figures/order_parameter_A_T_vs_degree.png",
    "/data4/toby/Branching-Search/runs/agents/e1_physics_dev/figures/order_parameter_A_T_vs_degree.pdf"
  ],
  "note": "A_t is the per-episode mean of the incorrect population summed over agents; the band is a 95% task-cluster bootstrap interval. endpoint_crossing solves log(A_T/A_0) = 0 and is a third convention, not the same estimator as the pooled lineage ratio in critical_degree.json: it includes lineages founded after round 1."
}