{
  "evaluated": "2026-07",
  "protocol": {
    "dataset": "LoCoMo",
    "questions": 1540,
    "categories": "1-4",
    "answerer": "gpt-4o-mini",
    "judge": "gpt-4o-mini",
    "temperature": 0,
    "metric": "J-score",
    "runs": 10,
    "frozen": true
  },
  "runs": [84.87, 85.84, 85.71, 85.58, 85.26, 85.78, 85.52, 86.17, 85.52, 85.19],
  "mean": 85.55,
  "std": 0.37,
  "comparisons": {
    "genesys": 85.55,
    "zep": 75.14,
    "mem0": 66.9
  },
  "disclosures": {
    "oracle_retrieval_ceiling": 94.9,
    "run_to_run_answer_churn": 0.054,
    "persistence_gate_g2p": {
      "gained": 170,
      "lost": 72,
      "ratio": "2.4:1",
      "verdict": "RED"
    },
    "non_comparable_modern_answerer": {
      "score": 89.68,
      "answerer": "gpt-4.1-mini",
      "comparable": false,
      "note": "Answerer swapped only; same memories, retrieval, and judge. Never presented as the headline."
    }
  },
  "source": "https://github.com/Astrix-Labs/genesys",
  "methodology": "https://genesys.astrixlabs.ai/developers/methodology"
}
