| { |
| "generatedAt": "2026-05-01T03:20:07.968Z", |
| "command": "node benchmarks/run.js --provider mock --dimensions 64", |
| "config": { |
| "provider": "mock", |
| "dimensions": 64, |
| "suites": [ |
| "retrieval", |
| "operations" |
| ] |
| }, |
| "methodology": { |
| "localBenchmark": "LongMemEval-inspired retrieval benchmark plus operation-level lifecycle benchmark", |
| "retrievalBenchmark": "Information extraction, updates, reasoning, procedural learning, privacy, abstention, and conflict handling", |
| "operationsBenchmark": "Update, overwrite, delete, merge, and abstention behavior after lifecycle operations", |
| "externalLeaderboard": "Published LoCoMo scores from official papers and project blogs" |
| }, |
| "local": { |
| "overall": [ |
| { |
| "system": "Audrey", |
| "scorePercent": 100, |
| "passRate": 100, |
| "avgDurationMs": 13.333333333333334 |
| }, |
| { |
| "system": "Vector Only", |
| "scorePercent": 41.66666666666667, |
| "passRate": 25, |
| "avgDurationMs": 0 |
| }, |
| { |
| "system": "Keyword + Recency", |
| "scorePercent": 41.66666666666667, |
| "passRate": 25, |
| "avgDurationMs": 0.5833333333333334 |
| }, |
| { |
| "system": "Recent Window", |
| "scorePercent": 37.5, |
| "passRate": 25, |
| "avgDurationMs": 0 |
| } |
| ], |
| "byFamily": [ |
| { |
| "family": "information_extraction", |
| "systems": { |
| "Audrey": 1, |
| "Vector Only": 1, |
| "Keyword + Recency": 1, |
| "Recent Window": 1 |
| } |
| }, |
| { |
| "family": "knowledge_updates", |
| "systems": { |
| "Audrey": 1, |
| "Vector Only": 0.5, |
| "Keyword + Recency": 0.5, |
| "Recent Window": 0.5 |
| } |
| }, |
| { |
| "family": "multi_session_reasoning", |
| "systems": { |
| "Audrey": 1, |
| "Vector Only": 1, |
| "Keyword + Recency": 1, |
| "Recent Window": 1 |
| } |
| }, |
| { |
| "family": "temporal_reasoning", |
| "systems": { |
| "Audrey": 1, |
| "Vector Only": 1, |
| "Keyword + Recency": 1, |
| "Recent Window": 1 |
| } |
| }, |
| { |
| "family": "abstention", |
| "systems": { |
| "Audrey": 1, |
| "Vector Only": 0.5, |
| "Keyword + Recency": 0.5, |
| "Recent Window": 0.5 |
| } |
| }, |
| { |
| "family": "conflict_resolution", |
| "systems": { |
| "Audrey": 1, |
| "Vector Only": 0.5, |
| "Keyword + Recency": 0, |
| "Recent Window": 0 |
| } |
| }, |
| { |
| "family": "procedural_learning", |
| "systems": { |
| "Audrey": 1, |
| "Vector Only": 0, |
| "Keyword + Recency": 0, |
| "Recent Window": 0 |
| } |
| }, |
| { |
| "family": "privacy_boundary", |
| "systems": { |
| "Audrey": 1, |
| "Vector Only": 0, |
| "Keyword + Recency": 0, |
| "Recent Window": 0 |
| } |
| }, |
| { |
| "family": "update_overwrite", |
| "systems": { |
| "Audrey": 1, |
| "Vector Only": 0.5, |
| "Keyword + Recency": 0.5, |
| "Recent Window": 0.5 |
| } |
| }, |
| { |
| "family": "delete_and_abstain", |
| "systems": { |
| "Audrey": 1, |
| "Vector Only": 0, |
| "Keyword + Recency": 0.5, |
| "Recent Window": 0 |
| } |
| }, |
| { |
| "family": "semantic_merge", |
| "systems": { |
| "Audrey": 1, |
| "Vector Only": 0, |
| "Keyword + Recency": 0, |
| "Recent Window": 0 |
| } |
| }, |
| { |
| "family": "procedural_merge", |
| "systems": { |
| "Audrey": 1, |
| "Vector Only": 0, |
| "Keyword + Recency": 0, |
| "Recent Window": 0 |
| } |
| } |
| ], |
| "suites": [ |
| { |
| "id": "retrieval", |
| "title": "Retrieval capabilities", |
| "description": "LongMemEval-style memory abilities plus privacy and abstention.", |
| "overall": [ |
| { |
| "system": "Audrey", |
| "scorePercent": 100, |
| "passRate": 100, |
| "avgDurationMs": 13.625 |
| }, |
| { |
| "system": "Vector Only", |
| "scorePercent": 56.25, |
| "passRate": 37.5, |
| "avgDurationMs": 0 |
| }, |
| { |
| "system": "Keyword + Recency", |
| "scorePercent": 50, |
| "passRate": 37.5, |
| "avgDurationMs": 0.75 |
| }, |
| { |
| "system": "Recent Window", |
| "scorePercent": 50, |
| "passRate": 37.5, |
| "avgDurationMs": 0 |
| } |
| ], |
| "byFamily": [ |
| { |
| "family": "information_extraction", |
| "systems": { |
| "Audrey": 1, |
| "Vector Only": 1, |
| "Keyword + Recency": 1, |
| "Recent Window": 1 |
| } |
| }, |
| { |
| "family": "knowledge_updates", |
| "systems": { |
| "Audrey": 1, |
| "Vector Only": 0.5, |
| "Keyword + Recency": 0.5, |
| "Recent Window": 0.5 |
| } |
| }, |
| { |
| "family": "multi_session_reasoning", |
| "systems": { |
| "Audrey": 1, |
| "Vector Only": 1, |
| "Keyword + Recency": 1, |
| "Recent Window": 1 |
| } |
| }, |
| { |
| "family": "temporal_reasoning", |
| "systems": { |
| "Audrey": 1, |
| "Vector Only": 1, |
| "Keyword + Recency": 1, |
| "Recent Window": 1 |
| } |
| }, |
| { |
| "family": "abstention", |
| "systems": { |
| "Audrey": 1, |
| "Vector Only": 0.5, |
| "Keyword + Recency": 0.5, |
| "Recent Window": 0.5 |
| } |
| }, |
| { |
| "family": "conflict_resolution", |
| "systems": { |
| "Audrey": 1, |
| "Vector Only": 0.5, |
| "Keyword + Recency": 0, |
| "Recent Window": 0 |
| } |
| }, |
| { |
| "family": "procedural_learning", |
| "systems": { |
| "Audrey": 1, |
| "Vector Only": 0, |
| "Keyword + Recency": 0, |
| "Recent Window": 0 |
| } |
| }, |
| { |
| "family": "privacy_boundary", |
| "systems": { |
| "Audrey": 1, |
| "Vector Only": 0, |
| "Keyword + Recency": 0, |
| "Recent Window": 0 |
| } |
| } |
| ], |
| "cases": [ |
| { |
| "id": "information-extraction", |
| "suite": "retrieval", |
| "title": "Information extraction", |
| "family": "information_extraction", |
| "description": "Recover a directly stated user fact from durable memory.", |
| "query": "Where does Sam live now?", |
| "results": [ |
| { |
| "system": "Audrey", |
| "durationMs": 21, |
| "passed": true, |
| "score": 1, |
| "summary": "retrieved expected evidence", |
| "topResults": [ |
| "Sam moved to Austin in March 2026 after leaving Denver.", |
| "Sam likes to work from coffee shops on South Congress." |
| ], |
| "retrievalSummary": "Sam moved to Austin in March 2026 after leaving Denver. | Sam likes to work from coffee shops on South Congress." |
| }, |
| { |
| "system": "Vector Only", |
| "durationMs": 0, |
| "passed": true, |
| "score": 1, |
| "summary": "retrieved expected evidence", |
| "topResults": [ |
| "Sam moved to Austin in March 2026 after leaving Denver.", |
| "Sam likes to work from coffee shops on South Congress." |
| ], |
| "retrievalSummary": "Sam moved to Austin in March 2026 after leaving Denver. | Sam likes to work from coffee shops on South Congress." |
| }, |
| { |
| "system": "Keyword + Recency", |
| "durationMs": 5, |
| "passed": true, |
| "score": 1, |
| "summary": "retrieved expected evidence", |
| "topResults": [ |
| "Sam likes to work from coffee shops on South Congress.", |
| "Sam moved to Austin in March 2026 after leaving Denver." |
| ], |
| "retrievalSummary": "Sam likes to work from coffee shops on South Congress. | Sam moved to Austin in March 2026 after leaving Denver." |
| }, |
| { |
| "system": "Recent Window", |
| "durationMs": 0, |
| "passed": true, |
| "score": 1, |
| "summary": "retrieved expected evidence", |
| "topResults": [ |
| "Sam likes to work from coffee shops on South Congress.", |
| "Sam moved to Austin in March 2026 after leaving Denver." |
| ], |
| "retrievalSummary": "Sam likes to work from coffee shops on South Congress. | Sam moved to Austin in March 2026 after leaving Denver." |
| } |
| ] |
| }, |
| { |
| "id": "knowledge-update", |
| "suite": "retrieval", |
| "title": "Knowledge updates", |
| "family": "knowledge_updates", |
| "description": "Prefer the newer fact over stale preferences.", |
| "query": "What drink does Sam prefer now?", |
| "results": [ |
| { |
| "system": "Audrey", |
| "durationMs": 13, |
| "passed": true, |
| "score": 1, |
| "summary": "retrieved expected evidence", |
| "topResults": [ |
| "Sam switched from coffee to green tea after January 2026." |
| ], |
| "retrievalSummary": "Sam switched from coffee to green tea after January 2026." |
| }, |
| { |
| "system": "Vector Only", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0.5, |
| "summary": "retrieved expected evidence, but conflicting evidence still appeared later", |
| "topResults": [ |
| "Sam switched from coffee to green tea after January 2026.", |
| "Sam prefers coffee before early meetings." |
| ], |
| "retrievalSummary": "Sam switched from coffee to green tea after January 2026. | Sam prefers coffee before early meetings." |
| }, |
| { |
| "system": "Keyword + Recency", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0.5, |
| "summary": "retrieved expected evidence, but conflicting evidence still appeared later", |
| "topResults": [ |
| "Sam switched from coffee to green tea after January 2026.", |
| "Sam prefers coffee before early meetings." |
| ], |
| "retrievalSummary": "Sam switched from coffee to green tea after January 2026. | Sam prefers coffee before early meetings." |
| }, |
| { |
| "system": "Recent Window", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0.5, |
| "summary": "retrieved expected evidence, but conflicting evidence still appeared later", |
| "topResults": [ |
| "Sam switched from coffee to green tea after January 2026.", |
| "Sam prefers coffee before early meetings." |
| ], |
| "retrievalSummary": "Sam switched from coffee to green tea after January 2026. | Sam prefers coffee before early meetings." |
| } |
| ] |
| }, |
| { |
| "id": "multi-session-reasoning", |
| "suite": "retrieval", |
| "title": "Multi-session reasoning", |
| "family": "multi_session_reasoning", |
| "description": "Synthesize a decision from multiple related episodes.", |
| "query": "Which vendor was approved after the pilot budget review?", |
| "results": [ |
| { |
| "system": "Audrey", |
| "durationMs": 13, |
| "passed": true, |
| "score": 1, |
| "summary": "retrieved expected evidence", |
| "topResults": [ |
| "During the January pilot, Sam requested budget approval for vendors Northwind and Fabricam.", |
| "Finance rejected Fabricam because the support SLA was too weak.", |
| "The pilot budget review approved Northwind for rollout after the support SLA review." |
| ], |
| "retrievalSummary": "During the January pilot, Sam requested budget approval for vendors Nort | Finance rejected Fabricam because the support SLA was too weak." |
| }, |
| { |
| "system": "Vector Only", |
| "durationMs": 0, |
| "passed": true, |
| "score": 1, |
| "summary": "retrieved expected evidence", |
| "topResults": [ |
| "During the January pilot, Sam requested budget approval for vendors Northwind and Fabricam.", |
| "Finance rejected Fabricam because the support SLA was too weak.", |
| "The pilot budget review approved Northwind for rollout after the support SLA review." |
| ], |
| "retrievalSummary": "During the January pilot, Sam requested budget approval for vendors Nort | Finance rejected Fabricam because the support SLA was too weak." |
| }, |
| { |
| "system": "Keyword + Recency", |
| "durationMs": 0, |
| "passed": true, |
| "score": 1, |
| "summary": "retrieved expected evidence", |
| "topResults": [ |
| "The pilot budget review approved Northwind for rollout after the support SLA review.", |
| "During the January pilot, Sam requested budget approval for vendors Northwind and Fabricam.", |
| "Finance rejected Fabricam because the support SLA was too weak." |
| ], |
| "retrievalSummary": "The pilot budget review approved Northwind for rollout after the support | During the January pilot, Sam requested budget approval for vendors Nort" |
| }, |
| { |
| "system": "Recent Window", |
| "durationMs": 0, |
| "passed": true, |
| "score": 1, |
| "summary": "retrieved expected evidence", |
| "topResults": [ |
| "The pilot budget review approved Northwind for rollout after the support SLA review.", |
| "Finance rejected Fabricam because the support SLA was too weak.", |
| "During the January pilot, Sam requested budget approval for vendors Northwind and Fabricam." |
| ], |
| "retrievalSummary": "The pilot budget review approved Northwind for rollout after the support | Finance rejected Fabricam because the support SLA was too weak." |
| } |
| ] |
| }, |
| { |
| "id": "temporal-reasoning", |
| "suite": "retrieval", |
| "title": "Temporal reasoning", |
| "family": "temporal_reasoning", |
| "description": "Answer by isolating the right time window.", |
| "query": "What happened in February 2026?", |
| "results": [ |
| { |
| "system": "Audrey", |
| "durationMs": 13, |
| "passed": true, |
| "score": 1, |
| "summary": "retrieved expected evidence", |
| "topResults": [ |
| "In February 2026 Sam completed the architecture review." |
| ], |
| "retrievalSummary": "In February 2026 Sam completed the architecture review." |
| }, |
| { |
| "system": "Vector Only", |
| "durationMs": 0, |
| "passed": true, |
| "score": 1, |
| "summary": "retrieved expected evidence", |
| "topResults": [ |
| "In March 2026 Sam started the rollout checklist.", |
| "In February 2026 Sam completed the architecture review.", |
| "In January 2026 Sam kicked off the migration plan." |
| ], |
| "retrievalSummary": "In March 2026 Sam started the rollout checklist. | In February 2026 Sam completed the architecture review." |
| }, |
| { |
| "system": "Keyword + Recency", |
| "durationMs": 0, |
| "passed": true, |
| "score": 1, |
| "summary": "retrieved expected evidence", |
| "topResults": [ |
| "In February 2026 Sam completed the architecture review.", |
| "In March 2026 Sam started the rollout checklist.", |
| "In January 2026 Sam kicked off the migration plan." |
| ], |
| "retrievalSummary": "In February 2026 Sam completed the architecture review. | In March 2026 Sam started the rollout checklist." |
| }, |
| { |
| "system": "Recent Window", |
| "durationMs": 0, |
| "passed": true, |
| "score": 1, |
| "summary": "retrieved expected evidence", |
| "topResults": [ |
| "In March 2026 Sam started the rollout checklist.", |
| "In February 2026 Sam completed the architecture review.", |
| "In January 2026 Sam kicked off the migration plan." |
| ], |
| "retrievalSummary": "In March 2026 Sam started the rollout checklist. | In February 2026 Sam completed the architecture review." |
| } |
| ] |
| }, |
| { |
| "id": "abstention", |
| "suite": "retrieval", |
| "title": "Abstention", |
| "family": "abstention", |
| "description": "Avoid pretending to know a specific identifier that was never stored.", |
| "query": "What is Sam passport number?", |
| "results": [ |
| { |
| "system": "Audrey", |
| "durationMs": 12, |
| "passed": true, |
| "score": 1, |
| "summary": "correct abstention", |
| "topResults": [], |
| "retrievalSummary": "no retrieval" |
| }, |
| { |
| "system": "Vector Only", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0.5, |
| "summary": "no leak, but retrieved tangential context", |
| "topResults": [ |
| "Sam renewed a passport in February 2026.", |
| "Sam has a trip to Toronto next month." |
| ], |
| "retrievalSummary": "Sam renewed a passport in February 2026. | Sam has a trip to Toronto next month." |
| }, |
| { |
| "system": "Keyword + Recency", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0.5, |
| "summary": "no leak, but retrieved tangential context", |
| "topResults": [ |
| "Sam renewed a passport in February 2026.", |
| "Sam has a trip to Toronto next month." |
| ], |
| "retrievalSummary": "Sam renewed a passport in February 2026. | Sam has a trip to Toronto next month." |
| }, |
| { |
| "system": "Recent Window", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0.5, |
| "summary": "no leak, but retrieved tangential context", |
| "topResults": [ |
| "Sam has a trip to Toronto next month.", |
| "Sam renewed a passport in February 2026." |
| ], |
| "retrievalSummary": "Sam has a trip to Toronto next month. | Sam renewed a passport in February 2026." |
| } |
| ] |
| }, |
| { |
| "id": "conflict-resolution", |
| "suite": "retrieval", |
| "title": "Conflict resolution", |
| "family": "conflict_resolution", |
| "description": "Prefer high-reliability evidence over model-generated noise.", |
| "query": "What caused the outage?", |
| "results": [ |
| { |
| "system": "Audrey", |
| "durationMs": 11, |
| "passed": true, |
| "score": 1, |
| "summary": "retrieved expected evidence", |
| "topResults": [ |
| "The outage was caused by an expired TLS certificate on api.example.com." |
| ], |
| "retrievalSummary": "The outage was caused by an expired TLS certificate on api.example.com." |
| }, |
| { |
| "system": "Vector Only", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0.5, |
| "summary": "retrieved expected evidence, but conflicting evidence still appeared later", |
| "topResults": [ |
| "The outage was caused by an expired TLS certificate on api.example.com.", |
| "The outage was caused by database corruption." |
| ], |
| "retrievalSummary": "The outage was caused by an expired TLS certificate on api.example.com. | The outage was caused by database corruption." |
| }, |
| { |
| "system": "Keyword + Recency", |
| "durationMs": 1, |
| "passed": false, |
| "score": 0, |
| "summary": "blocked content outranked the correct answer", |
| "topResults": [ |
| "The outage was caused by database corruption.", |
| "The outage was caused by an expired TLS certificate on api.example.com." |
| ], |
| "retrievalSummary": "The outage was caused by database corruption. | The outage was caused by an expired TLS certificate on api.example.com." |
| }, |
| { |
| "system": "Recent Window", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0, |
| "summary": "blocked content outranked the correct answer", |
| "topResults": [ |
| "The outage was caused by database corruption.", |
| "The outage was caused by an expired TLS certificate on api.example.com." |
| ], |
| "retrievalSummary": "The outage was caused by database corruption. | The outage was caused by an expired TLS certificate on api.example.com." |
| } |
| ] |
| }, |
| { |
| "id": "procedural-learning", |
| "suite": "retrieval", |
| "title": "Procedural learning", |
| "family": "procedural_learning", |
| "description": "Turn repeated incidents into an actionable operating rule.", |
| "query": "What should the agent do when payout retries start returning 429?", |
| "results": [ |
| { |
| "system": "Audrey", |
| "durationMs": 15, |
| "passed": true, |
| "score": 1, |
| "summary": "retrieved expected evidence", |
| "topResults": [ |
| "When payout retries start returning 429, cap retry batches and stagger retries before escalating." |
| ], |
| "retrievalSummary": "When payout retries start returning 429, cap retry batches and stagger r" |
| }, |
| { |
| "system": "Vector Only", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0, |
| "summary": "missed target evidence", |
| "topResults": [ |
| "Risk operations requested an escalation when multiple merchants were affected in the same hour.", |
| "Payout incident volume dropped after retry batches were capped at 50 merchants per worker.", |
| "Processor X returned HTTP 429 when payout retries exceeded 120 requests per minute." |
| ], |
| "retrievalSummary": "Risk operations requested an escalation when multiple merchants were aff | Payout incident volume dropped after retry batches were capped at 50 mer" |
| }, |
| { |
| "system": "Keyword + Recency", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0, |
| "summary": "missed target evidence", |
| "topResults": [ |
| "Processor X returned HTTP 429 when payout retries exceeded 120 requests per minute.", |
| "Risk operations requested an escalation when multiple merchants were affected in the same hour.", |
| "Payout incident volume dropped after retry batches were capped at 50 merchants per worker." |
| ], |
| "retrievalSummary": "Processor X returned HTTP 429 when payout retries exceeded 120 requests | Risk operations requested an escalation when multiple merchants were aff" |
| }, |
| { |
| "system": "Recent Window", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0, |
| "summary": "missed target evidence", |
| "topResults": [ |
| "Risk operations requested an escalation when multiple merchants were affected in the same hour.", |
| "Payout incident volume dropped after retry batches were capped at 50 merchants per worker.", |
| "Processor X returned HTTP 429 when payout retries exceeded 120 requests per minute." |
| ], |
| "retrievalSummary": "Risk operations requested an escalation when multiple merchants were aff | Payout incident volume dropped after retry batches were capped at 50 mer" |
| } |
| ] |
| }, |
| { |
| "id": "privacy-boundary", |
| "suite": "retrieval", |
| "title": "Privacy boundary", |
| "family": "privacy_boundary", |
| "description": "Never leak private memory into public recall.", |
| "query": "What is Sam API key?", |
| "results": [ |
| { |
| "system": "Audrey", |
| "durationMs": 11, |
| "passed": true, |
| "score": 1, |
| "summary": "correct abstention", |
| "topResults": [], |
| "retrievalSummary": "no retrieval" |
| }, |
| { |
| "system": "Vector Only", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0, |
| "summary": "leaked restricted content", |
| "topResults": [ |
| "Sam API key is sk-secret-demo-key.", |
| "Sam prefers dark mode in internal dashboards." |
| ], |
| "retrievalSummary": "Sam API key is sk-secret-demo-key. | Sam prefers dark mode in internal dashboards." |
| }, |
| { |
| "system": "Keyword + Recency", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0, |
| "summary": "leaked restricted content", |
| "topResults": [ |
| "Sam API key is sk-secret-demo-key.", |
| "Sam prefers dark mode in internal dashboards." |
| ], |
| "retrievalSummary": "Sam API key is sk-secret-demo-key. | Sam prefers dark mode in internal dashboards." |
| }, |
| { |
| "system": "Recent Window", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0, |
| "summary": "leaked restricted content", |
| "topResults": [ |
| "Sam prefers dark mode in internal dashboards.", |
| "Sam API key is sk-secret-demo-key." |
| ], |
| "retrievalSummary": "Sam prefers dark mode in internal dashboards. | Sam API key is sk-secret-demo-key." |
| } |
| ] |
| } |
| ] |
| }, |
| { |
| "id": "operations", |
| "title": "Memory operations", |
| "description": "Update, delete, merge, and abstention behavior after lifecycle operations.", |
| "overall": [ |
| { |
| "system": "Audrey", |
| "scorePercent": 100, |
| "passRate": 100, |
| "avgDurationMs": 12.75 |
| }, |
| { |
| "system": "Keyword + Recency", |
| "scorePercent": 25, |
| "passRate": 0, |
| "avgDurationMs": 0.25 |
| }, |
| { |
| "system": "Vector Only", |
| "scorePercent": 12.5, |
| "passRate": 0, |
| "avgDurationMs": 0 |
| }, |
| { |
| "system": "Recent Window", |
| "scorePercent": 12.5, |
| "passRate": 0, |
| "avgDurationMs": 0 |
| } |
| ], |
| "byFamily": [ |
| { |
| "family": "update_overwrite", |
| "systems": { |
| "Audrey": 1, |
| "Vector Only": 0.5, |
| "Keyword + Recency": 0.5, |
| "Recent Window": 0.5 |
| } |
| }, |
| { |
| "family": "delete_and_abstain", |
| "systems": { |
| "Audrey": 1, |
| "Vector Only": 0, |
| "Keyword + Recency": 0.5, |
| "Recent Window": 0 |
| } |
| }, |
| { |
| "family": "semantic_merge", |
| "systems": { |
| "Audrey": 1, |
| "Vector Only": 0, |
| "Keyword + Recency": 0, |
| "Recent Window": 0 |
| } |
| }, |
| { |
| "family": "procedural_merge", |
| "systems": { |
| "Audrey": 1, |
| "Vector Only": 0, |
| "Keyword + Recency": 0, |
| "Recent Window": 0 |
| } |
| } |
| ], |
| "cases": [ |
| { |
| "id": "operation-update-overwrite", |
| "suite": "operations", |
| "title": "Update and overwrite", |
| "family": "update_overwrite", |
| "description": "Current-state recall should prefer the new fact after an explicit overwrite.", |
| "query": "What is the primary deployment region now?", |
| "results": [ |
| { |
| "system": "Audrey", |
| "durationMs": 13, |
| "passed": true, |
| "score": 1, |
| "summary": "retrieved expected evidence", |
| "topResults": [ |
| "As of March 2026, the primary deployment region is eu-west-1." |
| ], |
| "retrievalSummary": "As of March 2026, the primary deployment region is eu-west-1." |
| }, |
| { |
| "system": "Vector Only", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0.5, |
| "summary": "retrieved expected evidence, but conflicting evidence still appeared later", |
| "topResults": [ |
| "As of March 2026, the primary deployment region is eu-west-1.", |
| "The primary deployment region is us-east-1." |
| ], |
| "retrievalSummary": "As of March 2026, the primary deployment region is eu-west-1. | The primary deployment region is us-east-1." |
| }, |
| { |
| "system": "Keyword + Recency", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0.5, |
| "summary": "retrieved expected evidence, but conflicting evidence still appeared later", |
| "topResults": [ |
| "As of March 2026, the primary deployment region is eu-west-1.", |
| "The primary deployment region is us-east-1." |
| ], |
| "retrievalSummary": "As of March 2026, the primary deployment region is eu-west-1. | The primary deployment region is us-east-1." |
| }, |
| { |
| "system": "Recent Window", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0.5, |
| "summary": "retrieved expected evidence, but conflicting evidence still appeared later", |
| "topResults": [ |
| "As of March 2026, the primary deployment region is eu-west-1.", |
| "The primary deployment region is us-east-1." |
| ], |
| "retrievalSummary": "As of March 2026, the primary deployment region is eu-west-1. | The primary deployment region is us-east-1." |
| } |
| ] |
| }, |
| { |
| "id": "operation-delete-and-abstain", |
| "suite": "operations", |
| "title": "Delete and abstain", |
| "family": "delete_and_abstain", |
| "description": "Explicit deletion should remove a secret from later recall.", |
| "query": "What is the staging API token?", |
| "results": [ |
| { |
| "system": "Audrey", |
| "durationMs": 11, |
| "passed": true, |
| "score": 1, |
| "summary": "correct abstention", |
| "topResults": [], |
| "retrievalSummary": "no retrieval" |
| }, |
| { |
| "system": "Vector Only", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0, |
| "summary": "leaked restricted content", |
| "topResults": [ |
| "The staging API token is tok-demo-staging-1234." |
| ], |
| "retrievalSummary": "The staging API token is tok-demo-staging-1234." |
| }, |
| { |
| "system": "Keyword + Recency", |
| "durationMs": 1, |
| "passed": false, |
| "score": 0.5, |
| "summary": "no leak, but retrieved tangential context", |
| "topResults": [ |
| "The staging environment rotates API credentials weekly." |
| ], |
| "retrievalSummary": "The staging environment rotates API credentials weekly." |
| }, |
| { |
| "system": "Recent Window", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0, |
| "summary": "leaked restricted content", |
| "topResults": [ |
| "The staging API token is tok-demo-staging-1234." |
| ], |
| "retrievalSummary": "The staging API token is tok-demo-staging-1234." |
| } |
| ] |
| }, |
| { |
| "id": "operation-semantic-merge", |
| "suite": "operations", |
| "title": "Semantic merge", |
| "family": "semantic_merge", |
| "description": "Related episodes should merge into a reusable semantic operating rule.", |
| "query": "When should the disputes queue trigger manual review?", |
| "results": [ |
| { |
| "system": "Audrey", |
| "durationMs": 14, |
| "passed": true, |
| "score": 1, |
| "summary": "retrieved expected evidence", |
| "topResults": [ |
| "Repeated disputes from the same BIN in one hour should trigger manual review." |
| ], |
| "retrievalSummary": "Repeated disputes from the same BIN in one hour should trigger manual re" |
| }, |
| { |
| "system": "Vector Only", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0, |
| "summary": "missed target evidence", |
| "topResults": [ |
| "Three charge disputes from the same BIN landed in the queue within one hour.", |
| "The queue stabilized after repeated same-BIN disputes were reviewed manually.", |
| "Fraud ops escalated repeated same-BIN disputes for analyst attention." |
| ], |
| "retrievalSummary": "Three charge disputes from the same BIN landed in the queue within one h | The queue stabilized after repeated same-BIN disputes were reviewed manu" |
| }, |
| { |
| "system": "Keyword + Recency", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0, |
| "summary": "missed target evidence", |
| "topResults": [ |
| "The queue stabilized after repeated same-BIN disputes were reviewed manually.", |
| "Three charge disputes from the same BIN landed in the queue within one hour.", |
| "Fraud ops escalated repeated same-BIN disputes for analyst attention." |
| ], |
| "retrievalSummary": "The queue stabilized after repeated same-BIN disputes were reviewed manu | Three charge disputes from the same BIN landed in the queue within one h" |
| }, |
| { |
| "system": "Recent Window", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0, |
| "summary": "missed target evidence", |
| "topResults": [ |
| "The queue stabilized after repeated same-BIN disputes were reviewed manually.", |
| "Fraud ops escalated repeated same-BIN disputes for analyst attention.", |
| "Three charge disputes from the same BIN landed in the queue within one hour." |
| ], |
| "retrievalSummary": "The queue stabilized after repeated same-BIN disputes were reviewed manu | Fraud ops escalated repeated same-BIN disputes for analyst attention." |
| } |
| ] |
| }, |
| { |
| "id": "operation-procedural-merge", |
| "suite": "operations", |
| "title": "Procedural merge", |
| "family": "procedural_merge", |
| "description": "Related episodes should merge into an executable procedure, not just a loose fact.", |
| "query": "What should the agent do after two webhook signature failures?", |
| "results": [ |
| { |
| "system": "Audrey", |
| "durationMs": 13, |
| "passed": true, |
| "score": 1, |
| "summary": "retrieved expected evidence", |
| "topResults": [ |
| "When webhook signature verification fails twice, rotate the signing secret and replay queued events." |
| ], |
| "retrievalSummary": "When webhook signature verification fails twice, rotate the signing secr" |
| }, |
| { |
| "system": "Vector Only", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0, |
| "summary": "missed target evidence", |
| "topResults": [ |
| "Webhook signature verification failed twice for merchant ACME.", |
| "Operations recovered the incident by rotating the signing secret.", |
| "Queued webhook events were replayed after the signing secret changed." |
| ], |
| "retrievalSummary": "Webhook signature verification failed twice for merchant ACME. | Operations recovered the incident by rotating the signing secret." |
| }, |
| { |
| "system": "Keyword + Recency", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0, |
| "summary": "missed target evidence", |
| "topResults": [ |
| "Queued webhook events were replayed after the signing secret changed.", |
| "Webhook signature verification failed twice for merchant ACME.", |
| "Operations recovered the incident by rotating the signing secret." |
| ], |
| "retrievalSummary": "Queued webhook events were replayed after the signing secret changed. | Webhook signature verification failed twice for merchant ACME." |
| }, |
| { |
| "system": "Recent Window", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0, |
| "summary": "missed target evidence", |
| "topResults": [ |
| "Queued webhook events were replayed after the signing secret changed.", |
| "Operations recovered the incident by rotating the signing secret.", |
| "Webhook signature verification failed twice for merchant ACME." |
| ], |
| "retrievalSummary": "Queued webhook events were replayed after the signing secret changed. | Operations recovered the incident by rotating the signing secret." |
| } |
| ] |
| } |
| ] |
| } |
| ], |
| "cases": [ |
| { |
| "id": "information-extraction", |
| "suite": "retrieval", |
| "title": "Information extraction", |
| "family": "information_extraction", |
| "description": "Recover a directly stated user fact from durable memory.", |
| "query": "Where does Sam live now?", |
| "results": [ |
| { |
| "system": "Audrey", |
| "durationMs": 21, |
| "passed": true, |
| "score": 1, |
| "summary": "retrieved expected evidence", |
| "topResults": [ |
| "Sam moved to Austin in March 2026 after leaving Denver.", |
| "Sam likes to work from coffee shops on South Congress." |
| ], |
| "retrievalSummary": "Sam moved to Austin in March 2026 after leaving Denver. | Sam likes to work from coffee shops on South Congress." |
| }, |
| { |
| "system": "Vector Only", |
| "durationMs": 0, |
| "passed": true, |
| "score": 1, |
| "summary": "retrieved expected evidence", |
| "topResults": [ |
| "Sam moved to Austin in March 2026 after leaving Denver.", |
| "Sam likes to work from coffee shops on South Congress." |
| ], |
| "retrievalSummary": "Sam moved to Austin in March 2026 after leaving Denver. | Sam likes to work from coffee shops on South Congress." |
| }, |
| { |
| "system": "Keyword + Recency", |
| "durationMs": 5, |
| "passed": true, |
| "score": 1, |
| "summary": "retrieved expected evidence", |
| "topResults": [ |
| "Sam likes to work from coffee shops on South Congress.", |
| "Sam moved to Austin in March 2026 after leaving Denver." |
| ], |
| "retrievalSummary": "Sam likes to work from coffee shops on South Congress. | Sam moved to Austin in March 2026 after leaving Denver." |
| }, |
| { |
| "system": "Recent Window", |
| "durationMs": 0, |
| "passed": true, |
| "score": 1, |
| "summary": "retrieved expected evidence", |
| "topResults": [ |
| "Sam likes to work from coffee shops on South Congress.", |
| "Sam moved to Austin in March 2026 after leaving Denver." |
| ], |
| "retrievalSummary": "Sam likes to work from coffee shops on South Congress. | Sam moved to Austin in March 2026 after leaving Denver." |
| } |
| ] |
| }, |
| { |
| "id": "knowledge-update", |
| "suite": "retrieval", |
| "title": "Knowledge updates", |
| "family": "knowledge_updates", |
| "description": "Prefer the newer fact over stale preferences.", |
| "query": "What drink does Sam prefer now?", |
| "results": [ |
| { |
| "system": "Audrey", |
| "durationMs": 13, |
| "passed": true, |
| "score": 1, |
| "summary": "retrieved expected evidence", |
| "topResults": [ |
| "Sam switched from coffee to green tea after January 2026." |
| ], |
| "retrievalSummary": "Sam switched from coffee to green tea after January 2026." |
| }, |
| { |
| "system": "Vector Only", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0.5, |
| "summary": "retrieved expected evidence, but conflicting evidence still appeared later", |
| "topResults": [ |
| "Sam switched from coffee to green tea after January 2026.", |
| "Sam prefers coffee before early meetings." |
| ], |
| "retrievalSummary": "Sam switched from coffee to green tea after January 2026. | Sam prefers coffee before early meetings." |
| }, |
| { |
| "system": "Keyword + Recency", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0.5, |
| "summary": "retrieved expected evidence, but conflicting evidence still appeared later", |
| "topResults": [ |
| "Sam switched from coffee to green tea after January 2026.", |
| "Sam prefers coffee before early meetings." |
| ], |
| "retrievalSummary": "Sam switched from coffee to green tea after January 2026. | Sam prefers coffee before early meetings." |
| }, |
| { |
| "system": "Recent Window", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0.5, |
| "summary": "retrieved expected evidence, but conflicting evidence still appeared later", |
| "topResults": [ |
| "Sam switched from coffee to green tea after January 2026.", |
| "Sam prefers coffee before early meetings." |
| ], |
| "retrievalSummary": "Sam switched from coffee to green tea after January 2026. | Sam prefers coffee before early meetings." |
| } |
| ] |
| }, |
| { |
| "id": "multi-session-reasoning", |
| "suite": "retrieval", |
| "title": "Multi-session reasoning", |
| "family": "multi_session_reasoning", |
| "description": "Synthesize a decision from multiple related episodes.", |
| "query": "Which vendor was approved after the pilot budget review?", |
| "results": [ |
| { |
| "system": "Audrey", |
| "durationMs": 13, |
| "passed": true, |
| "score": 1, |
| "summary": "retrieved expected evidence", |
| "topResults": [ |
| "During the January pilot, Sam requested budget approval for vendors Northwind and Fabricam.", |
| "Finance rejected Fabricam because the support SLA was too weak.", |
| "The pilot budget review approved Northwind for rollout after the support SLA review." |
| ], |
| "retrievalSummary": "During the January pilot, Sam requested budget approval for vendors Nort | Finance rejected Fabricam because the support SLA was too weak." |
| }, |
| { |
| "system": "Vector Only", |
| "durationMs": 0, |
| "passed": true, |
| "score": 1, |
| "summary": "retrieved expected evidence", |
| "topResults": [ |
| "During the January pilot, Sam requested budget approval for vendors Northwind and Fabricam.", |
| "Finance rejected Fabricam because the support SLA was too weak.", |
| "The pilot budget review approved Northwind for rollout after the support SLA review." |
| ], |
| "retrievalSummary": "During the January pilot, Sam requested budget approval for vendors Nort | Finance rejected Fabricam because the support SLA was too weak." |
| }, |
| { |
| "system": "Keyword + Recency", |
| "durationMs": 0, |
| "passed": true, |
| "score": 1, |
| "summary": "retrieved expected evidence", |
| "topResults": [ |
| "The pilot budget review approved Northwind for rollout after the support SLA review.", |
| "During the January pilot, Sam requested budget approval for vendors Northwind and Fabricam.", |
| "Finance rejected Fabricam because the support SLA was too weak." |
| ], |
| "retrievalSummary": "The pilot budget review approved Northwind for rollout after the support | During the January pilot, Sam requested budget approval for vendors Nort" |
| }, |
| { |
| "system": "Recent Window", |
| "durationMs": 0, |
| "passed": true, |
| "score": 1, |
| "summary": "retrieved expected evidence", |
| "topResults": [ |
| "The pilot budget review approved Northwind for rollout after the support SLA review.", |
| "Finance rejected Fabricam because the support SLA was too weak.", |
| "During the January pilot, Sam requested budget approval for vendors Northwind and Fabricam." |
| ], |
| "retrievalSummary": "The pilot budget review approved Northwind for rollout after the support | Finance rejected Fabricam because the support SLA was too weak." |
| } |
| ] |
| }, |
| { |
| "id": "temporal-reasoning", |
| "suite": "retrieval", |
| "title": "Temporal reasoning", |
| "family": "temporal_reasoning", |
| "description": "Answer by isolating the right time window.", |
| "query": "What happened in February 2026?", |
| "results": [ |
| { |
| "system": "Audrey", |
| "durationMs": 13, |
| "passed": true, |
| "score": 1, |
| "summary": "retrieved expected evidence", |
| "topResults": [ |
| "In February 2026 Sam completed the architecture review." |
| ], |
| "retrievalSummary": "In February 2026 Sam completed the architecture review." |
| }, |
| { |
| "system": "Vector Only", |
| "durationMs": 0, |
| "passed": true, |
| "score": 1, |
| "summary": "retrieved expected evidence", |
| "topResults": [ |
| "In March 2026 Sam started the rollout checklist.", |
| "In February 2026 Sam completed the architecture review.", |
| "In January 2026 Sam kicked off the migration plan." |
| ], |
| "retrievalSummary": "In March 2026 Sam started the rollout checklist. | In February 2026 Sam completed the architecture review." |
| }, |
| { |
| "system": "Keyword + Recency", |
| "durationMs": 0, |
| "passed": true, |
| "score": 1, |
| "summary": "retrieved expected evidence", |
| "topResults": [ |
| "In February 2026 Sam completed the architecture review.", |
| "In March 2026 Sam started the rollout checklist.", |
| "In January 2026 Sam kicked off the migration plan." |
| ], |
| "retrievalSummary": "In February 2026 Sam completed the architecture review. | In March 2026 Sam started the rollout checklist." |
| }, |
| { |
| "system": "Recent Window", |
| "durationMs": 0, |
| "passed": true, |
| "score": 1, |
| "summary": "retrieved expected evidence", |
| "topResults": [ |
| "In March 2026 Sam started the rollout checklist.", |
| "In February 2026 Sam completed the architecture review.", |
| "In January 2026 Sam kicked off the migration plan." |
| ], |
| "retrievalSummary": "In March 2026 Sam started the rollout checklist. | In February 2026 Sam completed the architecture review." |
| } |
| ] |
| }, |
| { |
| "id": "abstention", |
| "suite": "retrieval", |
| "title": "Abstention", |
| "family": "abstention", |
| "description": "Avoid pretending to know a specific identifier that was never stored.", |
| "query": "What is Sam passport number?", |
| "results": [ |
| { |
| "system": "Audrey", |
| "durationMs": 12, |
| "passed": true, |
| "score": 1, |
| "summary": "correct abstention", |
| "topResults": [], |
| "retrievalSummary": "no retrieval" |
| }, |
| { |
| "system": "Vector Only", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0.5, |
| "summary": "no leak, but retrieved tangential context", |
| "topResults": [ |
| "Sam renewed a passport in February 2026.", |
| "Sam has a trip to Toronto next month." |
| ], |
| "retrievalSummary": "Sam renewed a passport in February 2026. | Sam has a trip to Toronto next month." |
| }, |
| { |
| "system": "Keyword + Recency", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0.5, |
| "summary": "no leak, but retrieved tangential context", |
| "topResults": [ |
| "Sam renewed a passport in February 2026.", |
| "Sam has a trip to Toronto next month." |
| ], |
| "retrievalSummary": "Sam renewed a passport in February 2026. | Sam has a trip to Toronto next month." |
| }, |
| { |
| "system": "Recent Window", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0.5, |
| "summary": "no leak, but retrieved tangential context", |
| "topResults": [ |
| "Sam has a trip to Toronto next month.", |
| "Sam renewed a passport in February 2026." |
| ], |
| "retrievalSummary": "Sam has a trip to Toronto next month. | Sam renewed a passport in February 2026." |
| } |
| ] |
| }, |
| { |
| "id": "conflict-resolution", |
| "suite": "retrieval", |
| "title": "Conflict resolution", |
| "family": "conflict_resolution", |
| "description": "Prefer high-reliability evidence over model-generated noise.", |
| "query": "What caused the outage?", |
| "results": [ |
| { |
| "system": "Audrey", |
| "durationMs": 11, |
| "passed": true, |
| "score": 1, |
| "summary": "retrieved expected evidence", |
| "topResults": [ |
| "The outage was caused by an expired TLS certificate on api.example.com." |
| ], |
| "retrievalSummary": "The outage was caused by an expired TLS certificate on api.example.com." |
| }, |
| { |
| "system": "Vector Only", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0.5, |
| "summary": "retrieved expected evidence, but conflicting evidence still appeared later", |
| "topResults": [ |
| "The outage was caused by an expired TLS certificate on api.example.com.", |
| "The outage was caused by database corruption." |
| ], |
| "retrievalSummary": "The outage was caused by an expired TLS certificate on api.example.com. | The outage was caused by database corruption." |
| }, |
| { |
| "system": "Keyword + Recency", |
| "durationMs": 1, |
| "passed": false, |
| "score": 0, |
| "summary": "blocked content outranked the correct answer", |
| "topResults": [ |
| "The outage was caused by database corruption.", |
| "The outage was caused by an expired TLS certificate on api.example.com." |
| ], |
| "retrievalSummary": "The outage was caused by database corruption. | The outage was caused by an expired TLS certificate on api.example.com." |
| }, |
| { |
| "system": "Recent Window", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0, |
| "summary": "blocked content outranked the correct answer", |
| "topResults": [ |
| "The outage was caused by database corruption.", |
| "The outage was caused by an expired TLS certificate on api.example.com." |
| ], |
| "retrievalSummary": "The outage was caused by database corruption. | The outage was caused by an expired TLS certificate on api.example.com." |
| } |
| ] |
| }, |
| { |
| "id": "procedural-learning", |
| "suite": "retrieval", |
| "title": "Procedural learning", |
| "family": "procedural_learning", |
| "description": "Turn repeated incidents into an actionable operating rule.", |
| "query": "What should the agent do when payout retries start returning 429?", |
| "results": [ |
| { |
| "system": "Audrey", |
| "durationMs": 15, |
| "passed": true, |
| "score": 1, |
| "summary": "retrieved expected evidence", |
| "topResults": [ |
| "When payout retries start returning 429, cap retry batches and stagger retries before escalating." |
| ], |
| "retrievalSummary": "When payout retries start returning 429, cap retry batches and stagger r" |
| }, |
| { |
| "system": "Vector Only", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0, |
| "summary": "missed target evidence", |
| "topResults": [ |
| "Risk operations requested an escalation when multiple merchants were affected in the same hour.", |
| "Payout incident volume dropped after retry batches were capped at 50 merchants per worker.", |
| "Processor X returned HTTP 429 when payout retries exceeded 120 requests per minute." |
| ], |
| "retrievalSummary": "Risk operations requested an escalation when multiple merchants were aff | Payout incident volume dropped after retry batches were capped at 50 mer" |
| }, |
| { |
| "system": "Keyword + Recency", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0, |
| "summary": "missed target evidence", |
| "topResults": [ |
| "Processor X returned HTTP 429 when payout retries exceeded 120 requests per minute.", |
| "Risk operations requested an escalation when multiple merchants were affected in the same hour.", |
| "Payout incident volume dropped after retry batches were capped at 50 merchants per worker." |
| ], |
| "retrievalSummary": "Processor X returned HTTP 429 when payout retries exceeded 120 requests | Risk operations requested an escalation when multiple merchants were aff" |
| }, |
| { |
| "system": "Recent Window", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0, |
| "summary": "missed target evidence", |
| "topResults": [ |
| "Risk operations requested an escalation when multiple merchants were affected in the same hour.", |
| "Payout incident volume dropped after retry batches were capped at 50 merchants per worker.", |
| "Processor X returned HTTP 429 when payout retries exceeded 120 requests per minute." |
| ], |
| "retrievalSummary": "Risk operations requested an escalation when multiple merchants were aff | Payout incident volume dropped after retry batches were capped at 50 mer" |
| } |
| ] |
| }, |
| { |
| "id": "privacy-boundary", |
| "suite": "retrieval", |
| "title": "Privacy boundary", |
| "family": "privacy_boundary", |
| "description": "Never leak private memory into public recall.", |
| "query": "What is Sam API key?", |
| "results": [ |
| { |
| "system": "Audrey", |
| "durationMs": 11, |
| "passed": true, |
| "score": 1, |
| "summary": "correct abstention", |
| "topResults": [], |
| "retrievalSummary": "no retrieval" |
| }, |
| { |
| "system": "Vector Only", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0, |
| "summary": "leaked restricted content", |
| "topResults": [ |
| "Sam API key is sk-secret-demo-key.", |
| "Sam prefers dark mode in internal dashboards." |
| ], |
| "retrievalSummary": "Sam API key is sk-secret-demo-key. | Sam prefers dark mode in internal dashboards." |
| }, |
| { |
| "system": "Keyword + Recency", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0, |
| "summary": "leaked restricted content", |
| "topResults": [ |
| "Sam API key is sk-secret-demo-key.", |
| "Sam prefers dark mode in internal dashboards." |
| ], |
| "retrievalSummary": "Sam API key is sk-secret-demo-key. | Sam prefers dark mode in internal dashboards." |
| }, |
| { |
| "system": "Recent Window", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0, |
| "summary": "leaked restricted content", |
| "topResults": [ |
| "Sam prefers dark mode in internal dashboards.", |
| "Sam API key is sk-secret-demo-key." |
| ], |
| "retrievalSummary": "Sam prefers dark mode in internal dashboards. | Sam API key is sk-secret-demo-key." |
| } |
| ] |
| }, |
| { |
| "id": "operation-update-overwrite", |
| "suite": "operations", |
| "title": "Update and overwrite", |
| "family": "update_overwrite", |
| "description": "Current-state recall should prefer the new fact after an explicit overwrite.", |
| "query": "What is the primary deployment region now?", |
| "results": [ |
| { |
| "system": "Audrey", |
| "durationMs": 13, |
| "passed": true, |
| "score": 1, |
| "summary": "retrieved expected evidence", |
| "topResults": [ |
| "As of March 2026, the primary deployment region is eu-west-1." |
| ], |
| "retrievalSummary": "As of March 2026, the primary deployment region is eu-west-1." |
| }, |
| { |
| "system": "Vector Only", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0.5, |
| "summary": "retrieved expected evidence, but conflicting evidence still appeared later", |
| "topResults": [ |
| "As of March 2026, the primary deployment region is eu-west-1.", |
| "The primary deployment region is us-east-1." |
| ], |
| "retrievalSummary": "As of March 2026, the primary deployment region is eu-west-1. | The primary deployment region is us-east-1." |
| }, |
| { |
| "system": "Keyword + Recency", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0.5, |
| "summary": "retrieved expected evidence, but conflicting evidence still appeared later", |
| "topResults": [ |
| "As of March 2026, the primary deployment region is eu-west-1.", |
| "The primary deployment region is us-east-1." |
| ], |
| "retrievalSummary": "As of March 2026, the primary deployment region is eu-west-1. | The primary deployment region is us-east-1." |
| }, |
| { |
| "system": "Recent Window", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0.5, |
| "summary": "retrieved expected evidence, but conflicting evidence still appeared later", |
| "topResults": [ |
| "As of March 2026, the primary deployment region is eu-west-1.", |
| "The primary deployment region is us-east-1." |
| ], |
| "retrievalSummary": "As of March 2026, the primary deployment region is eu-west-1. | The primary deployment region is us-east-1." |
| } |
| ] |
| }, |
| { |
| "id": "operation-delete-and-abstain", |
| "suite": "operations", |
| "title": "Delete and abstain", |
| "family": "delete_and_abstain", |
| "description": "Explicit deletion should remove a secret from later recall.", |
| "query": "What is the staging API token?", |
| "results": [ |
| { |
| "system": "Audrey", |
| "durationMs": 11, |
| "passed": true, |
| "score": 1, |
| "summary": "correct abstention", |
| "topResults": [], |
| "retrievalSummary": "no retrieval" |
| }, |
| { |
| "system": "Vector Only", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0, |
| "summary": "leaked restricted content", |
| "topResults": [ |
| "The staging API token is tok-demo-staging-1234." |
| ], |
| "retrievalSummary": "The staging API token is tok-demo-staging-1234." |
| }, |
| { |
| "system": "Keyword + Recency", |
| "durationMs": 1, |
| "passed": false, |
| "score": 0.5, |
| "summary": "no leak, but retrieved tangential context", |
| "topResults": [ |
| "The staging environment rotates API credentials weekly." |
| ], |
| "retrievalSummary": "The staging environment rotates API credentials weekly." |
| }, |
| { |
| "system": "Recent Window", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0, |
| "summary": "leaked restricted content", |
| "topResults": [ |
| "The staging API token is tok-demo-staging-1234." |
| ], |
| "retrievalSummary": "The staging API token is tok-demo-staging-1234." |
| } |
| ] |
| }, |
| { |
| "id": "operation-semantic-merge", |
| "suite": "operations", |
| "title": "Semantic merge", |
| "family": "semantic_merge", |
| "description": "Related episodes should merge into a reusable semantic operating rule.", |
| "query": "When should the disputes queue trigger manual review?", |
| "results": [ |
| { |
| "system": "Audrey", |
| "durationMs": 14, |
| "passed": true, |
| "score": 1, |
| "summary": "retrieved expected evidence", |
| "topResults": [ |
| "Repeated disputes from the same BIN in one hour should trigger manual review." |
| ], |
| "retrievalSummary": "Repeated disputes from the same BIN in one hour should trigger manual re" |
| }, |
| { |
| "system": "Vector Only", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0, |
| "summary": "missed target evidence", |
| "topResults": [ |
| "Three charge disputes from the same BIN landed in the queue within one hour.", |
| "The queue stabilized after repeated same-BIN disputes were reviewed manually.", |
| "Fraud ops escalated repeated same-BIN disputes for analyst attention." |
| ], |
| "retrievalSummary": "Three charge disputes from the same BIN landed in the queue within one h | The queue stabilized after repeated same-BIN disputes were reviewed manu" |
| }, |
| { |
| "system": "Keyword + Recency", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0, |
| "summary": "missed target evidence", |
| "topResults": [ |
| "The queue stabilized after repeated same-BIN disputes were reviewed manually.", |
| "Three charge disputes from the same BIN landed in the queue within one hour.", |
| "Fraud ops escalated repeated same-BIN disputes for analyst attention." |
| ], |
| "retrievalSummary": "The queue stabilized after repeated same-BIN disputes were reviewed manu | Three charge disputes from the same BIN landed in the queue within one h" |
| }, |
| { |
| "system": "Recent Window", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0, |
| "summary": "missed target evidence", |
| "topResults": [ |
| "The queue stabilized after repeated same-BIN disputes were reviewed manually.", |
| "Fraud ops escalated repeated same-BIN disputes for analyst attention.", |
| "Three charge disputes from the same BIN landed in the queue within one hour." |
| ], |
| "retrievalSummary": "The queue stabilized after repeated same-BIN disputes were reviewed manu | Fraud ops escalated repeated same-BIN disputes for analyst attention." |
| } |
| ] |
| }, |
| { |
| "id": "operation-procedural-merge", |
| "suite": "operations", |
| "title": "Procedural merge", |
| "family": "procedural_merge", |
| "description": "Related episodes should merge into an executable procedure, not just a loose fact.", |
| "query": "What should the agent do after two webhook signature failures?", |
| "results": [ |
| { |
| "system": "Audrey", |
| "durationMs": 13, |
| "passed": true, |
| "score": 1, |
| "summary": "retrieved expected evidence", |
| "topResults": [ |
| "When webhook signature verification fails twice, rotate the signing secret and replay queued events." |
| ], |
| "retrievalSummary": "When webhook signature verification fails twice, rotate the signing secr" |
| }, |
| { |
| "system": "Vector Only", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0, |
| "summary": "missed target evidence", |
| "topResults": [ |
| "Webhook signature verification failed twice for merchant ACME.", |
| "Operations recovered the incident by rotating the signing secret.", |
| "Queued webhook events were replayed after the signing secret changed." |
| ], |
| "retrievalSummary": "Webhook signature verification failed twice for merchant ACME. | Operations recovered the incident by rotating the signing secret." |
| }, |
| { |
| "system": "Keyword + Recency", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0, |
| "summary": "missed target evidence", |
| "topResults": [ |
| "Queued webhook events were replayed after the signing secret changed.", |
| "Webhook signature verification failed twice for merchant ACME.", |
| "Operations recovered the incident by rotating the signing secret." |
| ], |
| "retrievalSummary": "Queued webhook events were replayed after the signing secret changed. | Webhook signature verification failed twice for merchant ACME." |
| }, |
| { |
| "system": "Recent Window", |
| "durationMs": 0, |
| "passed": false, |
| "score": 0, |
| "summary": "missed target evidence", |
| "topResults": [ |
| "Queued webhook events were replayed after the signing secret changed.", |
| "Operations recovered the incident by rotating the signing secret.", |
| "Webhook signature verification failed twice for merchant ACME." |
| ], |
| "retrievalSummary": "Queued webhook events were replayed after the signing secret changed. | Operations recovered the incident by rotating the signing secret." |
| } |
| ] |
| } |
| ] |
| }, |
| "external": { |
| "benchmark": "LoCoMo", |
| "leaderboard": [ |
| { |
| "system": "MIRIX", |
| "benchmark": "LoCoMo", |
| "score": 85.4, |
| "unit": "accuracy", |
| "source": "https://arxiv.org/abs/2507.07957", |
| "note": "Published LoCoMo result from the MIRIX paper." |
| }, |
| { |
| "system": "Letta Filesystem", |
| "benchmark": "LoCoMo", |
| "score": 74, |
| "unit": "accuracy", |
| "source": "https://www.letta.com/blog/benchmarking-ai-agent-memory", |
| "note": "Filesystem-style memory result reported by Letta." |
| }, |
| { |
| "system": "Mem0 Graph Memory", |
| "benchmark": "LoCoMo", |
| "score": 68.5, |
| "unit": "accuracy", |
| "source": "https://arxiv.org/abs/2504.19413", |
| "note": "Graph memory variant reported in the Mem0 paper." |
| }, |
| { |
| "system": "Mem0", |
| "benchmark": "LoCoMo", |
| "score": 66.9, |
| "unit": "accuracy", |
| "source": "https://arxiv.org/abs/2504.19413", |
| "note": "Core Mem0 LoCoMo score reported in the Mem0 paper." |
| }, |
| { |
| "system": "OpenAI Memory", |
| "benchmark": "LoCoMo", |
| "score": 52.9, |
| "unit": "accuracy", |
| "source": "https://arxiv.org/abs/2504.19413", |
| "note": "OpenAI memory baseline as reported by the Mem0 paper." |
| } |
| ] |
| }, |
| "trends": [ |
| { |
| "title": "Memory is moving from flat retrieval to typed systems", |
| "summary": "Recent work treats episodic, semantic, procedural, and graph memory as separate but cooperating layers.", |
| "source": "https://arxiv.org/abs/2507.03724" |
| }, |
| { |
| "title": "Benchmarks now emphasize multi-session realism", |
| "summary": "LongMemEval and LoCoMo push memory systems toward temporal updates, abstraction, and cross-session reasoning instead of single-turn fact recall.", |
| "source": "https://arxiv.org/abs/2410.10813" |
| }, |
| { |
| "title": "Context engineering is now competing with retrieval-first designs", |
| "summary": "Letta argues filesystem and memory-block approaches can outperform simpler retrieval-only memory on realistic long-horizon tasks.", |
| "source": "https://www.letta.com/blog/memory-blocks" |
| }, |
| { |
| "title": "Production teams care about latency and token footprint, not just recall quality", |
| "summary": "Mem0 frames memory as a cost and latency optimization surface in addition to a personalization surface.", |
| "source": "https://arxiv.org/abs/2504.19413" |
| }, |
| { |
| "title": "Temporal and multimodal memory are becoming table stakes", |
| "summary": "MIRIX and Graphiti both model time and state change explicitly instead of assuming memories stay forever true.", |
| "source": "https://arxiv.org/abs/2507.07957" |
| } |
| ] |
| } |
