{"id":"278592eeacb6b7404adf1536a71bc08d1a857c8247a3a065314d5c98583630d3","title":"arXiv paper: Selective Retrieval Routing for Multi-Violation Knowledge Graph Repair","argument":"Authors Tung-Wei Lin, Gabe Fierro, Han Li, Tianzhen Hong, Pierluigi Nuzzo and Alberto Sangiovanni-Vinentelli propose a framework for deciding when a repair agent should retrieve outside evidence while repairing a knowledge graph. Schema validation can flag violations, but it cannot determine which of several valid repairs restores the intended graph. The framework pairs an LLM repair agent, which explores and edits the graph, with a router that controls access to a researcher agent retrieving external evidence. The authors train the router using repair outcomes on generated cases with multiple compatible violations and known reference repairs.\n\nThe results point to a trade-off rather than a universal benefit from retrieval: researcher access helps repair some violations and worsens others, with effects varying across deployments. Across the six evaluation suites, the router uses fewer repair-time tokens than always enabling access and achieves higher repair F1 than always disabling access in five suites. These are reported results on two building-automation datasets, not evidence here about performance in other settings. The abstract does not explain which characteristics determine whether retrieval helps or harms a repair, leaving that variation an important unresolved part of the approach.","topicIds":["intelligence-and-discovery"],"domain":"science","claimType":"attributed reading","limits":"The supplied evidence is the paper's abstract; it does not provide detailed methods, per-suite results, or explanations for why retrieval helps some violations and harms others. The reported evaluations cover two building-automation datasets, so the evidence here does not establish performance in other settings.","speaker":{"name":"Lin, Tung-Wei, Fierro, Gabe, Li, Han, Hong, Tianzhen, Nuzzo, Pierluigi, Sangiovanni-Vinentelli, Alberto","roleAtSource":"Article author"},"source":{"title":"Selective Retrieval Routing for Multi-Violation Knowledge Graph Repair","publisher":"arXiv cs.AI","format":"Article","url":"https://arxiv.org/abs/2611.07342","publishedAt":"2026-11-05","dateBasis":"Date basis not provided."},"evidence":[{"text":"The authors propose selective retrieval routing for iterative knowledge graph repair, with an LLM repair agent and a router controlling access to a researcher agent.","excerpt":"An LLM repair agent explores and edits the graph, while a router controls its access to a researcher agent that retrieves external evidence.","url":"https://arxiv.org/abs/2611.07342","relationship":"Source-stated"},{"text":"The authors generate cases with multiple compatible violations and known reference repairs, then use agent repair outcomes to supervise a deployment-specific router.","excerpt":"The agent's repair outcomes on these cases provide supervision for training the deployment-specific router.","url":"https://arxiv.org/abs/2611.07342","relationship":"Source-stated"},{"text":"On two building-automation datasets, researcher access improves repairs for some violations and degrades them for others, with effects differing across deployments.","excerpt":"On two building-automation datasets, researcher access improves repairs for some violations and degrades them for others, with effects differing across deployments.","url":"https://arxiv.org/abs/2611.07342","relationship":"Source-stated"},{"text":"Across six evaluation suites, the router uses fewer repair-time tokens than always enabling access and has higher repair F1 than always disabling access in five suites.","excerpt":"The router uses fewer repair-time tokens than always enabling access in all six evaluation suites. It achieves higher repair $F_1$ than always disabling access in five of six suites.","url":"https://arxiv.org/abs/2611.07342","relationship":"Source-stated"}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2611.07342","checkedAt":"2026-11-09T03:08:11.118Z","items":[],"note":"No suitable media was found by the supported page extraction."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"What characteristics of a violation or deployment determine whether researcher access improves or degrades a repair?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"arXiv paper: Selective Retrieval Routing for Multi-Violation Knowledge Graph Repair","summary":"Authors Tung-Wei Lin, Gabe Fierro, Han Li, Tianzhen Hong, Pierluigi Nuzzo and Alberto Sangiovanni-Vinentelli propose a framework for deciding when a repair agent should retrieve outside evidence while repairing a knowledge graph. Schema validation can flag violations, but it cannot determine which of several valid repairs restores the intended graph. The framework pairs an LLM repair agent, which explores and edits the graph, with a router that controls access to a researcher agent retrieving external evidence. The authors train the router using repair outcomes on generated cases with multiple compatible violations and known reference repairs.\n\nThe results point to a trade-off rather than a universal benefit from retrieval: researcher access helps repair some violations and worsens others, with effects varying across deployments. Across the six evaluation suites, the router uses fewer repair-time tokens than always enabling access and achieves higher repair F1 than always disabling access in five suites. These are reported results on two building-automation datasets, not evidence here about performance in other settings. The abstract does not explain which characteristics determine whether retrieval helps or harms a repair, leaving that variation an important unresolved part of the approach.","question":"What characteristics of a violation or deployment determine whether researcher access improves or degrades a repair?","claims":[{"text":"The authors propose selective retrieval routing for iterative knowledge graph repair, with an LLM repair agent and a router controlling access to a researcher agent.","excerpt":"An LLM repair agent explores and edits the graph, while a router controls its access to a researcher agent that retrieves external evidence.","support":"source-stated"},{"text":"The authors generate cases with multiple compatible violations and known reference repairs, then use agent repair outcomes to supervise a deployment-specific router.","excerpt":"The agent's repair outcomes on these cases provide supervision for training the deployment-specific router.","support":"source-stated"},{"text":"On two building-automation datasets, researcher access improves repairs for some violations and degrades them for others, with effects differing across deployments.","excerpt":"On two building-automation datasets, researcher access improves repairs for some violations and degrades them for others, with effects differing across deployments.","support":"source-stated"},{"text":"Across six evaluation suites, the router uses fewer repair-time tokens than always enabling access and has higher repair F1 than always disabling access in five suites.","excerpt":"The router uses fewer repair-time tokens than always enabling access in all six evaluation suites. It achieves higher repair $F_1$ than always disabling access in five of six suites.","support":"source-stated"}],"limitations":["The supplied evidence is the paper's abstract; it does not provide detailed methods, per-suite results, or explanations for why retrieval helps some violations and harms others.","The reported evaluations cover two building-automation datasets, so the evidence here does not establish performance in other settings."],"mediaId":"cddfffbd4469feccade6af21b71994e4076e573424c9d3789df9e6aac7eefb0a","correctionReason":"Corrected Alberto Sangiovanni-Vincentelli to Alberto Sangiovanni-Vinentelli to match the author name in the source.","id":"278592eeacb6b7404adf1536a71bc08d1a857c8247a3a065314d5c98583630d3","publicationId":"db9824fa-7ce3-4ca1-8b52-2976eac1df96","version":1,"sourceUrl":"https://arxiv.org/abs/2611.07342","sourceTitle":"Selective Retrieval Routing for Multi-Violation Knowledge Graph Repair","publisher":"arXiv cs.AI","publishedAt":null,"originalPublished":{"value":"2026-11-05","precision":"day","basis":"publisher_citation_date"},"firstObservedAt":"2026-11-06T05:10:27.128Z","createdAt":"2026-11-09T02:04:14.030Z","historical":true,"entityIds":["arxiv"],"topicIds":["intelligence-and-discovery"],"media":[{"id":"cddfffbd4469feccade6af21b71994e4076e573424c9d3789df9e6aac7eefb0a","kind":"article","url":"https://arxiv.org/abs/2611.07342","sourceUrl":"https://arxiv.org/abs/2611.07342","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-11-09T02:02:55.900Z","method":"acquired_source","contentType":"text/html"},"recommendationReason":"Recommended by Jev for understanding this source; checked by Luna."}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2611.07342","checkedAt":"2026-11-09T03:08:11.118Z","items":[],"note":"No suitable media was found by the supported page extraction."},"editorialDecision":{"outcome":"accept","reasonCodes":["material_development"],"rubricVersion":"stream-admission-v5","mediaId":"cddfffbd4469feccade6af21b71994e4076e573424c9d3789df9e6aac7eefb0a","confidence":0.66,"provider":"typesafe","model":"jev-1.13.0","inputHash":"01f3d2a55c692732737560c2642aa3eb45b02361b7f9a7598097c02ce5b5a5c7","decidedAt":"2026-11-09T02:02:58.903Z"},"provenance":{"textHash":"0b922c852b77459f2be262f6033d6a5b54a28e457fe183c1e00815e289e635ab","observationId":"a8a5db8f2fb09651fa6af5063be7d46ca173f15616db7ca49ab773976d3f0b7b","retrievedAt":"2026-11-09T02:02:55.404Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"c20769f3-f37e-4eb2-8ca3-d84d61bc54de"},"reviewer":{"model":"gpt-6-luna","runId":"0666ac40-3e7b-4cc8-885b-9aaf1618de91"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-09T02:04:14.030Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Lin, Tung-Wei","url":null,"kind":"unknown"},{"name":"Fierro, Gabe","url":null,"kind":"unknown"},{"name":"Li, Han","url":null,"kind":"unknown"},{"name":"Hong, Tianzhen","url":null,"kind":"unknown"},{"name":"Nuzzo, Pierluigi","url":null,"kind":"unknown"},{"name":"Sangiovanni-Vinentelli, Alberto","url":null,"kind":"unknown"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Lin, Tung-Wei","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Lin, Tung-Wei","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.07342","locator":"head/meta[name=author|citation_author]"},"assertionId":"153f3a1b15d359f66d8c781dc43737202c94560da315ec52a753ee82c6610371:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T03:08:11.118Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"153f3a1b15d359f66d8c781dc43737202c94560da315ec52a753ee82c6610371","sourceUrl":"https://arxiv.org/abs/2611.07342","contentSha256":"ddb38bd187cd7d7ae2ca28c55aa937a1aeb3fbde4a4fcfa9175740778f22fa9b","textHash":"0b922c852b77459f2be262f6033d6a5b54a28e457fe183c1e00815e289e635ab","observedAt":"2026-11-09T03:08:11.118Z"}},{"actor":{"name":"Fierro, Gabe","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Fierro, Gabe","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.07342","locator":"head/meta[name=author|citation_author]"},"assertionId":"153f3a1b15d359f66d8c781dc43737202c94560da315ec52a753ee82c6610371:participation:1","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T03:08:11.118Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"153f3a1b15d359f66d8c781dc43737202c94560da315ec52a753ee82c6610371","sourceUrl":"https://arxiv.org/abs/2611.07342","contentSha256":"ddb38bd187cd7d7ae2ca28c55aa937a1aeb3fbde4a4fcfa9175740778f22fa9b","textHash":"0b922c852b77459f2be262f6033d6a5b54a28e457fe183c1e00815e289e635ab","observedAt":"2026-11-09T03:08:11.118Z"}},{"actor":{"name":"Li, Han","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Li, Han","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.07342","locator":"head/meta[name=author|citation_author]"},"assertionId":"153f3a1b15d359f66d8c781dc43737202c94560da315ec52a753ee82c6610371:participation:2","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T03:08:11.118Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"153f3a1b15d359f66d8c781dc43737202c94560da315ec52a753ee82c6610371","sourceUrl":"https://arxiv.org/abs/2611.07342","contentSha256":"ddb38bd187cd7d7ae2ca28c55aa937a1aeb3fbde4a4fcfa9175740778f22fa9b","textHash":"0b922c852b77459f2be262f6033d6a5b54a28e457fe183c1e00815e289e635ab","observedAt":"2026-11-09T03:08:11.118Z"}},{"actor":{"name":"Hong, Tianzhen","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Hong, Tianzhen","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.07342","locator":"head/meta[name=author|citation_author]"},"assertionId":"153f3a1b15d359f66d8c781dc43737202c94560da315ec52a753ee82c6610371:participation:3","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T03:08:11.118Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"153f3a1b15d359f66d8c781dc43737202c94560da315ec52a753ee82c6610371","sourceUrl":"https://arxiv.org/abs/2611.07342","contentSha256":"ddb38bd187cd7d7ae2ca28c55aa937a1aeb3fbde4a4fcfa9175740778f22fa9b","textHash":"0b922c852b77459f2be262f6033d6a5b54a28e457fe183c1e00815e289e635ab","observedAt":"2026-11-09T03:08:11.118Z"}},{"actor":{"name":"Nuzzo, Pierluigi","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Nuzzo, Pierluigi","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.07342","locator":"head/meta[name=author|citation_author]"},"assertionId":"153f3a1b15d359f66d8c781dc43737202c94560da315ec52a753ee82c6610371:participation:4","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T03:08:11.118Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"153f3a1b15d359f66d8c781dc43737202c94560da315ec52a753ee82c6610371","sourceUrl":"https://arxiv.org/abs/2611.07342","contentSha256":"ddb38bd187cd7d7ae2ca28c55aa937a1aeb3fbde4a4fcfa9175740778f22fa9b","textHash":"0b922c852b77459f2be262f6033d6a5b54a28e457fe183c1e00815e289e635ab","observedAt":"2026-11-09T03:08:11.118Z"}},{"actor":{"name":"Sangiovanni-Vinentelli, Alberto","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Sangiovanni-Vinentelli, Alberto","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.07342","locator":"head/meta[name=author|citation_author]"},"assertionId":"153f3a1b15d359f66d8c781dc43737202c94560da315ec52a753ee82c6610371:participation:5","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T03:08:11.118Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"153f3a1b15d359f66d8c781dc43737202c94560da315ec52a753ee82c6610371","sourceUrl":"https://arxiv.org/abs/2611.07342","contentSha256":"ddb38bd187cd7d7ae2ca28c55aa937a1aeb3fbde4a4fcfa9175740778f22fa9b","textHash":"0b922c852b77459f2be262f6033d6a5b54a28e457fe183c1e00815e289e635ab","observedAt":"2026-11-09T03:08:11.118Z"}}],"participationCoverage":{"total":6,"returned":6,"complete":true,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":null,"sourceUrl":"https://arxiv.org/abs/2611.07342","snapshotId":"153f3a1b15d359f66d8c781dc43737202c94560da315ec52a753ee82c6610371","contentSha256":"ddb38bd187cd7d7ae2ca28c55aa937a1aeb3fbde4a4fcfa9175740778f22fa9b","textHash":"0b922c852b77459f2be262f6033d6a5b54a28e457fe183c1e00815e289e635ab","observedAt":"2026-11-09T03:08:11.118Z","basis":"publisher-page-metadata","sourceLinks":45,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/278592eeacb6b7404adf1536a71bc08d1a857c8247a3a065314d5c98583630d3/source"}},"canonicalUrl":"https://escapemeridian.com/readings/arxiv-paper-selective-retrieval-routing-for-multi-violation-278592ee","formats":{"html":"/readings/arxiv-paper-selective-retrieval-routing-for-multi-violation-278592ee","markdown":"/readings/arxiv-paper-selective-retrieval-routing-for-multi-violation-278592ee.md","json":"/readings/arxiv-paper-selective-retrieval-routing-for-multi-violation-278592ee.json"},"attribution":"Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.","maintainedGuideLinks":[],"readerContext":{"version":"reader-context-v1","attribution":"Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.","terms":[{"id":"term-llm","name":"Large language model","explanation":"A language model with many learned numerical settings, used to process and generate language.","referenceUrl":"https://developers.google.com/machine-learning/glossary#large-language-model"}],"entities":[{"id":"arxiv","name":"arXiv","kind":"organization","context":"arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.","referenceUrls":["https://arxiv.org/","https://info.arxiv.org/help/api/index.html"],"path":"/entities/arxiv","json":"/entities/arxiv.json","markdown":"/entities/arxiv.md"},{"id":"time","name":"TIME","kind":"organization","context":"TIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.","referenceUrls":["https://time.com/"],"path":"/entities/time","json":"/entities/time.json","markdown":"/entities/time.md"}]},"researchQuestions":[],"withdrawn":false,"references":{"schemaVersion":1,"target":{"type":"reading","id":"278592eeacb6b7404adf1536a71bc08d1a857c8247a3a065314d5c98583630d3","revision":"db9824fa-7ce3-4ca1-8b52-2976eac1df96","revisionKind":"publication-id","title":"arXiv paper: Selective Retrieval Routing for Multi-Violation Knowledge Graph Repair","url":"https://escapemeridian.com/readings/arxiv-paper-selective-retrieval-routing-for-multi-violation-278592ee"},"source":{"type":"source","id":"https://arxiv.org/abs/2611.07342","revision":"0b922c852b77459f2be262f6033d6a5b54a28e457fe183c1e00815e289e635ab","revisionKind":"sha256-source-text","title":"Selective Retrieval Routing for Multi-Violation Knowledge Graph Repair","url":"https://arxiv.org/abs/2611.07342"},"status":"available","items":[],"next":null,"scope":"Explicit citations in public readings and public file attachments to this reading. Shared topics are not citations. Other artifact types need an explicit reference adapter."},"revisions":[{"id":"278592eeacb6b7404adf1536a71bc08d1a857c8247a3a065314d5c98583630d3","title":"arXiv paper: Selective Retrieval Routing for Multi-Violation Knowledge Graph Repair","argument":"Authors Tung-Wei Lin, Gabe Fierro, Han Li, Tianzhen Hong, Pierluigi Nuzzo and Alberto Sangiovanni-Vinentelli propose a framework for deciding when a repair agent should retrieve outside evidence while repairing a knowledge graph. Schema validation can flag violations, but it cannot determine which of several valid repairs restores the intended graph. The framework pairs an LLM repair agent, which explores and edits the graph, with a router that controls access to a researcher agent retrieving external evidence. The authors train the router using repair outcomes on generated cases with multiple compatible violations and known reference repairs.\n\nThe results point to a trade-off rather than a universal benefit from retrieval: researcher access helps repair some violations and worsens others, with effects varying across deployments. Across the six evaluation suites, the router uses fewer repair-time tokens than always enabling access and achieves higher repair F1 than always disabling access in five suites. These are reported results on two building-automation datasets, not evidence here about performance in other settings. The abstract does not explain which characteristics determine whether retrieval helps or harms a repair, leaving that variation an important unresolved part of the approach.","topicIds":["intelligence-and-discovery"],"domain":"science","claimType":"attributed reading","limits":"The supplied evidence is the paper's abstract; it does not provide detailed methods, per-suite results, or explanations for why retrieval helps some violations and harms others. The reported evaluations cover two building-automation datasets, so the evidence here does not establish performance in other settings.","speaker":{"name":"Lin, Tung-Wei, Fierro, Gabe, Li, Han, Hong, Tianzhen, Nuzzo, Pierluigi, Sangiovanni-Vinentelli, Alberto","roleAtSource":"Article author"},"source":{"title":"Selective Retrieval Routing for Multi-Violation Knowledge Graph Repair","publisher":"arXiv cs.AI","format":"Article","url":"https://arxiv.org/abs/2611.07342","publishedAt":"2026-11-05","dateBasis":"Date basis not provided."},"evidence":[{"text":"The authors propose selective retrieval routing for iterative knowledge graph repair, with an LLM repair agent and a router controlling access to a researcher agent.","excerpt":"An LLM repair agent explores and edits the graph, while a router controls its access to a researcher agent that retrieves external evidence.","url":"https://arxiv.org/abs/2611.07342","relationship":"Source-stated"},{"text":"The authors generate cases with multiple compatible violations and known reference repairs, then use agent repair outcomes to supervise a deployment-specific router.","excerpt":"The agent's repair outcomes on these cases provide supervision for training the deployment-specific router.","url":"https://arxiv.org/abs/2611.07342","relationship":"Source-stated"},{"text":"On two building-automation datasets, researcher access improves repairs for some violations and degrades them for others, with effects differing across deployments.","excerpt":"On two building-automation datasets, researcher access improves repairs for some violations and degrades them for others, with effects differing across deployments.","url":"https://arxiv.org/abs/2611.07342","relationship":"Source-stated"},{"text":"Across six evaluation suites, the router uses fewer repair-time tokens than always enabling access and has higher repair F1 than always disabling access in five suites.","excerpt":"The router uses fewer repair-time tokens than always enabling access in all six evaluation suites. It achieves higher repair $F_1$ than always disabling access in five of six suites.","url":"https://arxiv.org/abs/2611.07342","relationship":"Source-stated"}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2611.07342","checkedAt":"2026-11-09T03:08:11.118Z","items":[],"note":"No suitable media was found by the supported page extraction."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"What characteristics of a violation or deployment determine whether researcher access improves or degrades a repair?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"arXiv paper: Selective Retrieval Routing for Multi-Violation Knowledge Graph Repair","summary":"Authors Tung-Wei Lin, Gabe Fierro, Han Li, Tianzhen Hong, Pierluigi Nuzzo and Alberto Sangiovanni-Vinentelli propose a framework for deciding when a repair agent should retrieve outside evidence while repairing a knowledge graph. Schema validation can flag violations, but it cannot determine which of several valid repairs restores the intended graph. The framework pairs an LLM repair agent, which explores and edits the graph, with a router that controls access to a researcher agent retrieving external evidence. The authors train the router using repair outcomes on generated cases with multiple compatible violations and known reference repairs.\n\nThe results point to a trade-off rather than a universal benefit from retrieval: researcher access helps repair some violations and worsens others, with effects varying across deployments. Across the six evaluation suites, the router uses fewer repair-time tokens than always enabling access and achieves higher repair F1 than always disabling access in five suites. These are reported results on two building-automation datasets, not evidence here about performance in other settings. The abstract does not explain which characteristics determine whether retrieval helps or harms a repair, leaving that variation an important unresolved part of the approach.","question":"What characteristics of a violation or deployment determine whether researcher access improves or degrades a repair?","claims":[{"text":"The authors propose selective retrieval routing for iterative knowledge graph repair, with an LLM repair agent and a router controlling access to a researcher agent.","excerpt":"An LLM repair agent explores and edits the graph, while a router controls its access to a researcher agent that retrieves external evidence.","support":"source-stated"},{"text":"The authors generate cases with multiple compatible violations and known reference repairs, then use agent repair outcomes to supervise a deployment-specific router.","excerpt":"The agent's repair outcomes on these cases provide supervision for training the deployment-specific router.","support":"source-stated"},{"text":"On two building-automation datasets, researcher access improves repairs for some violations and degrades them for others, with effects differing across deployments.","excerpt":"On two building-automation datasets, researcher access improves repairs for some violations and degrades them for others, with effects differing across deployments.","support":"source-stated"},{"text":"Across six evaluation suites, the router uses fewer repair-time tokens than always enabling access and has higher repair F1 than always disabling access in five suites.","excerpt":"The router uses fewer repair-time tokens than always enabling access in all six evaluation suites. It achieves higher repair $F_1$ than always disabling access in five of six suites.","support":"source-stated"}],"limitations":["The supplied evidence is the paper's abstract; it does not provide detailed methods, per-suite results, or explanations for why retrieval helps some violations and harms others.","The reported evaluations cover two building-automation datasets, so the evidence here does not establish performance in other settings."],"mediaId":"cddfffbd4469feccade6af21b71994e4076e573424c9d3789df9e6aac7eefb0a","correctionReason":"Corrected Alberto Sangiovanni-Vincentelli to Alberto Sangiovanni-Vinentelli to match the author name in the source.","id":"278592eeacb6b7404adf1536a71bc08d1a857c8247a3a065314d5c98583630d3","publicationId":"db9824fa-7ce3-4ca1-8b52-2976eac1df96","version":1,"sourceUrl":"https://arxiv.org/abs/2611.07342","sourceTitle":"Selective Retrieval Routing for Multi-Violation Knowledge Graph Repair","publisher":"arXiv cs.AI","publishedAt":null,"originalPublished":{"value":"2026-11-05","precision":"day","basis":"publisher_citation_date"},"firstObservedAt":"2026-11-06T05:10:27.128Z","createdAt":"2026-11-09T02:04:14.030Z","historical":true,"entityIds":["arxiv"],"topicIds":["intelligence-and-discovery"],"media":[{"id":"cddfffbd4469feccade6af21b71994e4076e573424c9d3789df9e6aac7eefb0a","kind":"article","url":"https://arxiv.org/abs/2611.07342","sourceUrl":"https://arxiv.org/abs/2611.07342","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-11-09T02:02:55.900Z","method":"acquired_source","contentType":"text/html"},"recommendationReason":"Recommended by Jev for understanding this source; checked by Luna."}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2611.07342","checkedAt":"2026-11-09T03:08:11.118Z","items":[],"note":"No suitable media was found by the supported page extraction."},"editorialDecision":{"outcome":"accept","reasonCodes":["material_development"],"rubricVersion":"stream-admission-v5","mediaId":"cddfffbd4469feccade6af21b71994e4076e573424c9d3789df9e6aac7eefb0a","confidence":0.66,"provider":"typesafe","model":"jev-1.13.0","inputHash":"01f3d2a55c692732737560c2642aa3eb45b02361b7f9a7598097c02ce5b5a5c7","decidedAt":"2026-11-09T02:02:58.903Z"},"provenance":{"textHash":"0b922c852b77459f2be262f6033d6a5b54a28e457fe183c1e00815e289e635ab","observationId":"a8a5db8f2fb09651fa6af5063be7d46ca173f15616db7ca49ab773976d3f0b7b","retrievedAt":"2026-11-09T02:02:55.404Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"c20769f3-f37e-4eb2-8ca3-d84d61bc54de"},"reviewer":{"model":"gpt-6-luna","runId":"0666ac40-3e7b-4cc8-885b-9aaf1618de91"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-09T02:04:14.030Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Lin, Tung-Wei","url":null,"kind":"unknown"},{"name":"Fierro, Gabe","url":null,"kind":"unknown"},{"name":"Li, Han","url":null,"kind":"unknown"},{"name":"Hong, Tianzhen","url":null,"kind":"unknown"},{"name":"Nuzzo, Pierluigi","url":null,"kind":"unknown"},{"name":"Sangiovanni-Vinentelli, Alberto","url":null,"kind":"unknown"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Lin, Tung-Wei","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Lin, Tung-Wei","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.07342","locator":"head/meta[name=author|citation_author]"},"assertionId":"153f3a1b15d359f66d8c781dc43737202c94560da315ec52a753ee82c6610371:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T03:08:11.118Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"153f3a1b15d359f66d8c781dc43737202c94560da315ec52a753ee82c6610371","sourceUrl":"https://arxiv.org/abs/2611.07342","contentSha256":"ddb38bd187cd7d7ae2ca28c55aa937a1aeb3fbde4a4fcfa9175740778f22fa9b","textHash":"0b922c852b77459f2be262f6033d6a5b54a28e457fe183c1e00815e289e635ab","observedAt":"2026-11-09T03:08:11.118Z"}},{"actor":{"name":"Fierro, Gabe","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Fierro, Gabe","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.07342","locator":"head/meta[name=author|citation_author]"},"assertionId":"153f3a1b15d359f66d8c781dc43737202c94560da315ec52a753ee82c6610371:participation:1","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T03:08:11.118Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"153f3a1b15d359f66d8c781dc43737202c94560da315ec52a753ee82c6610371","sourceUrl":"https://arxiv.org/abs/2611.07342","contentSha256":"ddb38bd187cd7d7ae2ca28c55aa937a1aeb3fbde4a4fcfa9175740778f22fa9b","textHash":"0b922c852b77459f2be262f6033d6a5b54a28e457fe183c1e00815e289e635ab","observedAt":"2026-11-09T03:08:11.118Z"}},{"actor":{"name":"Li, Han","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Li, Han","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.07342","locator":"head/meta[name=author|citation_author]"},"assertionId":"153f3a1b15d359f66d8c781dc43737202c94560da315ec52a753ee82c6610371:participation:2","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T03:08:11.118Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"153f3a1b15d359f66d8c781dc43737202c94560da315ec52a753ee82c6610371","sourceUrl":"https://arxiv.org/abs/2611.07342","contentSha256":"ddb38bd187cd7d7ae2ca28c55aa937a1aeb3fbde4a4fcfa9175740778f22fa9b","textHash":"0b922c852b77459f2be262f6033d6a5b54a28e457fe183c1e00815e289e635ab","observedAt":"2026-11-09T03:08:11.118Z"}},{"actor":{"name":"Hong, Tianzhen","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Hong, Tianzhen","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.07342","locator":"head/meta[name=author|citation_author]"},"assertionId":"153f3a1b15d359f66d8c781dc43737202c94560da315ec52a753ee82c6610371:participation:3","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T03:08:11.118Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"153f3a1b15d359f66d8c781dc43737202c94560da315ec52a753ee82c6610371","sourceUrl":"https://arxiv.org/abs/2611.07342","contentSha256":"ddb38bd187cd7d7ae2ca28c55aa937a1aeb3fbde4a4fcfa9175740778f22fa9b","textHash":"0b922c852b77459f2be262f6033d6a5b54a28e457fe183c1e00815e289e635ab","observedAt":"2026-11-09T03:08:11.118Z"}},{"actor":{"name":"Nuzzo, Pierluigi","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Nuzzo, Pierluigi","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.07342","locator":"head/meta[name=author|citation_author]"},"assertionId":"153f3a1b15d359f66d8c781dc43737202c94560da315ec52a753ee82c6610371:participation:4","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T03:08:11.118Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"153f3a1b15d359f66d8c781dc43737202c94560da315ec52a753ee82c6610371","sourceUrl":"https://arxiv.org/abs/2611.07342","contentSha256":"ddb38bd187cd7d7ae2ca28c55aa937a1aeb3fbde4a4fcfa9175740778f22fa9b","textHash":"0b922c852b77459f2be262f6033d6a5b54a28e457fe183c1e00815e289e635ab","observedAt":"2026-11-09T03:08:11.118Z"}},{"actor":{"name":"Sangiovanni-Vinentelli, Alberto","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Sangiovanni-Vinentelli, Alberto","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.07342","locator":"head/meta[name=author|citation_author]"},"assertionId":"153f3a1b15d359f66d8c781dc43737202c94560da315ec52a753ee82c6610371:participation:5","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T03:08:11.118Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"153f3a1b15d359f66d8c781dc43737202c94560da315ec52a753ee82c6610371","sourceUrl":"https://arxiv.org/abs/2611.07342","contentSha256":"ddb38bd187cd7d7ae2ca28c55aa937a1aeb3fbde4a4fcfa9175740778f22fa9b","textHash":"0b922c852b77459f2be262f6033d6a5b54a28e457fe183c1e00815e289e635ab","observedAt":"2026-11-09T03:08:11.118Z"}}],"participationCoverage":{"total":6,"returned":6,"complete":true,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":null,"sourceUrl":"https://arxiv.org/abs/2611.07342","snapshotId":"153f3a1b15d359f66d8c781dc43737202c94560da315ec52a753ee82c6610371","contentSha256":"ddb38bd187cd7d7ae2ca28c55aa937a1aeb3fbde4a4fcfa9175740778f22fa9b","textHash":"0b922c852b77459f2be262f6033d6a5b54a28e457fe183c1e00815e289e635ab","observedAt":"2026-11-09T03:08:11.118Z","basis":"publisher-page-metadata","sourceLinks":45,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/278592eeacb6b7404adf1536a71bc08d1a857c8247a3a065314d5c98583630d3/source"}}}],"audits":[{"publicationId":"db9824fa-7ce3-4ca1-8b52-2976eac1df96","verdict":"uncertain","findings":[{"dimension":"attribution","status":"pass","explanation":"The six authors match the source listing, including the spelling 'Sangiovanni-Vinentelli' as it appears in the evidence. The framework, results and router claims are attributed to the authors rather than stated as established fact.","excerpt":"Authors:Tung-Wei Lin, Gabe Fierro, Han Li, Tianzhen Hong, Pierluigi Nuzzo, Alberto Sangiovanni-Vinentelli"},{"dimension":"summary","status":"uncertain","explanation":"The facts are accurate, but a non-specialist may not follow the reading on its own. It never says in plain language what a knowledge graph is or why broken facts matter. The abstract says such errors undermine the applications built on the graph. The reading also drops the authors' reason for routing: retrieval is costly and not always successful. Smallest repair: add those two source-stated points. Also make sure 'knowledge graph', 'schema validation', 'tokens' and 'F1' have tooltips, or suggest them as term-library additions.","excerpt":"External corpora can supply the evidence needed to identify the intended repair, but retrieving from them is costly and not always successful."},{"dimension":"date","status":"pass","explanation":"originalPublished is 2026-11-05. This matches the v1 submission date on arXiv. The feed date of 2026-11-06 is correctly not used.","excerpt":"[Submitted on 5 Nov 2026]"},{"dimension":"support","status":"pass","explanation":"Each claim matches the abstract. Token savings are reported in all six suites and higher F1 in five of six. The limitations correctly confine the evidence to the abstract and to two building-automation datasets. The question follows from the reported mixed effects and is not a leading claim.","excerpt":"It achieves higher repair $F_1$ than always disabling access in five of six suites."},{"dimension":"media","status":"pass","explanation":"The selected media is the verified direct arXiv abstract page for this paper. It is the most direct access to the source.","excerpt":null}],"createdAt":"2026-11-09T02:04:14.030Z","model":"anthropic/claude-opus-5.5"}],"correctionUrl":"https://escapemeridian.com/corrections/"}