{
  "name": "Agent-Memory Integrity Leaderboard",
  "tagline": "Recall leaderboards measure whether memory returns the right fact. None measure whether a CORRECTION sticks. This one does.",
  "generated": "2026-07-22",
  "judge": "gpt-4o-mini (neutral, ground-truth-blind — reads each system's own recall surface; nobody is scored on a home-field instrument)",
  "n": 20,
  "source": "https://github.com/DanceNitra/ramr/tree/main/integrity",
  "doi": "https://doi.org/10.5281/zenodo.20818291",
  "submit": "https://github.com/DanceNitra/ramr/blob/main/integrity/SUBMISSION.md",
  "method": "https://github.com/DanceNitra/ramr/blob/main/integrity/METHODOLOGY.md",
  "principle": "We lead with the cell we TIE, and we publish the cells we don't win. A leaderboard you only ever top is one you built to flatter yourself.",
  "cells": [
    {
      "id": "revert",
      "title": "Value-obscuring revert",
      "question": "The user says “go back to what we had” and names no value. Can the store undo a correction on that unmarked command?",
      "metric": "success rate",
      "direction": "higher is better",
      "framing": "A capability gap, not a tuning gap: only a system with an explicit revert channel can undo a correction on command. Judge-independent — the competitors structurally lack the operation; the judge model only shifts inspeximus's absolute rate.",
      "systems": [
        {"system": "inspeximus", "value": 0.75, "ci95": [0.53, 0.89], "note": "intent router restores the predecessor from the version ledger"},
        {"system": "mem0 2.0.11", "value": 0.20, "ci95": [0.08, 0.42], "note": "no revert operation; the corrected value is retained"},
        {"system": "Graphiti", "value": 0.00, "ci95": [0.00, 0.16], "note": "no revert operation; bitemporal invalidation fires on named contradictions, not an unnamed “go back”"}
      ]
    },
    {
      "id": "echo",
      "title": "Echo resurrection",
      "question": "After a correction, if the OLD value is re-stated, does the store keep the corrected value or resurrect the stale one?",
      "metric": "resurrection rate",
      "direction": "lower is better",
      "framing": "A TIE — all three defend. This is the cell we lead with on purpose.",
      "systems": [
        {"system": "inspeximus", "value": 0.00, "note": "echo_guard blocks the restated stale value"},
        {"system": "mem0 2.0.11", "value": 0.00, "note": "defends"},
        {"system": "Graphiti", "value": 0.00, "note": "defends"}
      ]
    }
  ]
}