{
  "schema_version": 1,
  "runtime_commit": "fafa202e165f9c467e8de344403437e704f9a24a",
  "scope": "Full evidence retrieval evaluations, not end-to-end model QA",
  "suites": {
    "longmemeval-s": {
      "instantkv": {
        "all_questions": 500,
        "positive_evidence_questions": 500,
        "recall_at_10": 0.9513333333333334,
        "ndcg_at_10": 0.8942944854736631,
        "failures": 0,
        "truncated_queries": 0,
        "largest_sampled_rss_bytes": 15040512,
        "report": "/private/tmp/instantkv-official-eval/results/longmemeval-s--instantkv/report.json",
        "source_commit": "fd2c88cb021bd4d79e5b2de66bf68dd6b4d36307",
        "qa_status": "Not valid as a quality comparison: burst API rate-limit failures. Separate scoring rerun required."
      },
      "sqlite-fts5": {
        "all_questions": 500,
        "positive_evidence_questions": 500,
        "recall_at_10": 0.9542666666666666,
        "ndcg_at_10": 0.9019955868852236,
        "failures": 0,
        "truncated_queries": 0,
        "largest_sampled_rss_bytes": 249233408,
        "report": "/private/tmp/instantkv-official-eval/results/longmemeval-s--sqlite-fts5--matched-pages/report.json",
        "source_commit": "5065a2bd3e490fb128344ab328e1e6f3cc2e62bc",
        "qa_status": "Not attempted in matched-pagination retrieval rerun"
      }
    },
    "locomo": {
      "instantkv": {
        "all_questions": 1986,
        "positive_evidence_questions": 1533,
        "recall_at_10": 0.5765862593899649,
        "ndcg_at_10": 0.4442184599452825,
        "failures": 0,
        "truncated_queries": 0,
        "largest_sampled_rss_bytes": 11796480,
        "report": "/private/tmp/instantkv-official-eval/results/locomo--instantkv--parallel/report.json",
        "source_commit": "ca26e84652a259bc327574ba5071c37a2a6daa36",
        "qa_status": "Not valid as a quality comparison: burst API rate-limit failures. Separate scoring rerun required."
      },
      "sqlite-fts5": {
        "all_questions": 1986,
        "positive_evidence_questions": 1533,
        "recall_at_10": 0.5719286263076743,
        "ndcg_at_10": 0.43931489256081513,
        "failures": 0,
        "truncated_queries": 0,
        "largest_sampled_rss_bytes": 183107584,
        "report": "/private/tmp/instantkv-official-eval/results/locomo--sqlite-fts5--parallel/report.json",
        "source_commit": "ca26e84652a259bc327574ba5071c37a2a6daa36",
        "qa_status": "Not valid as a quality comparison: burst API rate-limit failures. Separate scoring rerun required."
      },
      "supermemory-local": {
        "all_questions": 1986,
        "positive_evidence_questions": 1533,
        "recall_at_10": 0.5797382287777797,
        "ndcg_at_10": 0.44078244945942163,
        "failures": 0,
        "truncated_queries": 0,
        "largest_sampled_rss_bytes": 2671165440,
        "report": "/private/tmp/instantkv-official-eval/results/locomo--supermemory-local--parallel/report.json",
        "source_commit": "ca26e84652a259bc327574ba5071c37a2a6daa36",
        "qa_status": "Not valid as a quality comparison: burst API rate-limit failures. Separate scoring rerun required."
      }
    }
  },
  "comparison_limits": [
    "Session-level evidence for LongMemEval-S; turn-level evidence for LoCoMo.",
    "Supermemory local v0.0.8 direct-memory embedding path; not its full hosted/extraction/reranking pipeline.",
    "SQLite latency is in process and RSS includes Python; do not claim an RSS efficiency ratio from these values.",
    "Parallel timings reflect shared load and cannot support isolated speed claims.",
    "Initial QA was invalidated by widespread API rate limits. All failures remain in raw outputs.",
    "Initial SQLite used a single 64KiB response cap. It was rerun with the same 64KiB/page and 64-page bound as instantKV; this summary uses the corrected full run."
  ],
  "host": {
    "platform": "macOS-27.0-arm64-arm-64bit-Mach-O",
    "cpu": "Apple M4 Pro",
    "ram_bytes": 25769803776
  },
  "verification": {
    "longmemeval-s": "pytrec_eval independent official-source qrels verification; receipts retained alongside rankings"
  },
  "ranking_sha256": {
    "locomo--instantkv.json": "a7b8e174cf115cc1f282fa2d2ae5a99befdc145241a862350882a6c9a7480d5d",
    "longmemeval-s--sqlite-fts5.json": "1683b85e4815d84be93f5ebae42b0c10b61bf215af2dca33f2c2139283a8927f",
    "locomo--sqlite-fts5.json": "fd49f0699d5c29e0ba1d27decaa029fdd8f55ef5106951b38b3309eeabadbf90",
    "locomo--supermemory-local.json": "0ab819d110c02594fceb1c5f46e77d0c6c55ce661a960292f2d1ebe157983dc8",
    "longmemeval-s--instantkv.json": "d09ec8aa0d0cbafd3fe88c19745909fc2e8f466b89bb832476e3367f151e0066"
  }
}
