{
  "title": "Mnemosyne Recall \u2014 recorded evaluation",
  "release": "2.0.0-rc.8 source release candidate",
  "protocol": "mnemosyne-corpus-retrieval-v1",
  "dataset": {
    "sourceUrl": "https://huggingface.co/datasets/xiaowu0162/longmemeval-cleaned/resolve/98d7416c24c778c2fee6e6f3006e7a073259d48f/longmemeval_s_cleaned.json",
    "revision": "98d7416c24c778c2fee6e6f3006e7a073259d48f",
    "license": "MIT",
    "sourceSha256": "d6f21ea9d60a0d56f34a05b609c79c88a451d2ae03597821ea3d5a9678c3a442",
    "sourceBytes": 277383467
  },
  "reportSha256": "508d78ffd585e9af75359c377c6bd88cca6bcfc660714a92704571008152ef1a",
  "measuredImplementationSha256": "781e7e538a49910d25673c3d8b1f8f5b40e29b912f57195417fe3a16e4d0eac3",
  "questions": 500,
  "positiveAnnotatedQuestions": 470,
  "abstentionQuestions": 30,
  "conditions": [
    {
      "name": "Previous lexical scoring",
      "allSessionsRetrieved": 386,
      "allSessionsInPackedContext": 367
    },
    {
      "name": "Mnemosyne upgraded lexical scoring",
      "allSessionsRetrieved": 425,
      "allSessionsInPackedContext": 382
    }
  ],
  "settings": {
    "retrievedChunks": 20,
    "sourceChunkBytes": 4096,
    "contextEnvelopeBytes": 8192
  },
  "calls": {
    "embedding": 0,
    "embeddingInputBytes": 0,
    "generation": 0,
    "judge": 0
  },
  "answerQuality": "not-evaluated",
  "limits": [
    "Session-level source coverage does not prove answer-bearing passage coverage or answer correctness.",
    "Question-day compatibility includes 1475 source sessions later than their question instant on the same day.",
    "All 500 full-context baselines exceeded the context envelope.",
    "Frozen project build; later provider-only dependency validation did not change measured retrieval or evaluation code."
  ],
  "lifecycleDiagnostic": {
    "scenarios": 16,
    "before": {
      "passed": 9,
      "failed": 1,
      "errors": 6
    },
    "after": {
      "passed": 15,
      "failed": 1,
      "errors": 0
    },
    "openCase": "Unrelated source-field edits can conservatively retire valid dependent guidance. Alternative sufficient evidence is not supported."
  },
  "dataPolicy": "Aggregate measurements only. No source conversations, private memory, credentials or local filesystem paths are included.",
  "publication": {
    "githubRelease": "https://github.com/28naem-del/mnemosyne/releases/tag/v2.0.0-rc.9",
    "measuredSourceRelease": "https://github.com/28naem-del/mnemosyne/releases/tag/v2.0.0-rc.8",
    "npm": {
      "published": true,
      "package": "mnemosy-ai",
      "version": "2.0.0-rc.9",
      "distTag": "next",
      "latestVersion": "1.0.1",
      "url": "https://www.npmjs.com/package/mnemosy-ai/v/2.0.0-rc.9"
    },
    "pypi": {
      "published": false,
      "distribution": "mnemosyne-memory-client",
      "availability": "source checkout"
    },
    "note": "Publication status is separate from the frozen rc.8 evaluation above. The rc.9 packaging release does not represent a new benchmark run."
  }
}