{"id":"3783528233dd2f0e437395c96b0985ca9ba91f235fd191349fc9ac3243e0f86a","title":"arXiv: Route to Remember: Adaptive Memory Selection for LLM Agents","argument":"The authors of “Route to Remember: Adaptive Memory Selection for LLM Agents” argue that weight memory could help agents retain facts without adding text to every query. Weight memory stores knowledge in model parameters; the paper pairs it with text memory, which retrieves written information into the model’s context. Their method uses on-policy self-distillation (OPSD): a context-free student learns from a teacher model given the user’s memory. A verify-then-regenerate router selects between weight and text memory for each query. On the HaluMem benchmark, the authors report that OPSD matches or surpasses strong text-memory baselines, including full-context prompting and retrieval systems, while reading fewer tokens per query. That result suggests a possible way to reduce query-time reading while retaining access to user-specific facts, but the abstract does not establish performance in deployed agents. The authors also study how different training recipes affect weight memory. They present it as a practical, low-cost component for persistent, multi-session agents, while identifying training and evaluation protocols for combining it with retrieval as an area their work informs. The evidence here is an abstract: it does not give benchmark details or establish how the reported results transfer beyond HaluMem.","topicIds":["intelligence-and-discovery"],"domain":"science","claimType":"attributed reading","limits":"The available evidence is the arXiv abstract; it does not include benchmark details, experimental setup, or numerical results. The abstract reports results on HaluMem but does not establish whether they transfer to other factual-memory evaluations or deployed agents.","speaker":{"name":"Fang, Zeyu, Yin, Jun, Mei, Yongsheng, Han, Xu, Joshi, Chaitali, Sun, Zhongkai, Hao, Jie, Guo, Bin, Ma, Chengyuan, Yao, Benjamin Z., Guo, Chenlei","roleAtSource":"Article author"},"source":{"title":"Route to Remember: Adaptive Memory Selection for LLM Agents","publisher":"arXiv cs.AI","format":"Article","url":"https://arxiv.org/abs/2611.05636","publishedAt":"2026-11-03","dateBasis":"Date basis not provided."},"evidence":[{"text":"The authors propose combining weight memory with text memory using a router that chooses which memory handles a query.","excerpt":"a verify-then-regenerate router that lets weight and text memory each cover the queries they handle best","url":"https://arxiv.org/abs/2611.05636","relationship":"Source-stated"},{"text":"The authors report that OPSD matches or surpasses strong text-memory baselines on HaluMem while reading fewer tokens per query.","excerpt":"On the HaluMem benchmark, OPSD matches or surpasses strong text-memory baselines, including full-context prompting and retrieval systems, while reading fewer tokens per query.","url":"https://arxiv.org/abs/2611.05636","relationship":"Source-stated"},{"text":"The authors present weight memory as a potential low-cost component for agents with persistent, multi-session use.","excerpt":"these results position weight memory as a practical, low-cost component of agent memory","url":"https://arxiv.org/abs/2611.05636","relationship":"Source-stated"},{"text":"The abstract describes a study of how training recipes affect weight memory's performance.","excerpt":"We further conduct an extensive study that characterizes the behavior of weight memory and analyzes how different training recipes affect its performance.","url":"https://arxiv.org/abs/2611.05636","relationship":"Source-stated"}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2611.05636","checkedAt":"2026-11-09T11:52:10.428Z","items":[],"note":"No suitable media was found by the supported page extraction."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"How does OPSD perform on factual-memory evaluations beyond HaluMem, compared with the text-memory baselines?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"arXiv: Route to Remember: Adaptive Memory Selection for LLM Agents","summary":"The authors of “Route to Remember: Adaptive Memory Selection for LLM Agents” argue that weight memory could help agents retain facts without adding text to every query. Weight memory stores knowledge in model parameters; the paper pairs it with text memory, which retrieves written information into the model’s context. Their method uses on-policy self-distillation (OPSD): a context-free student learns from a teacher model given the user’s memory. A verify-then-regenerate router selects between weight and text memory for each query. On the HaluMem benchmark, the authors report that OPSD matches or surpasses strong text-memory baselines, including full-context prompting and retrieval systems, while reading fewer tokens per query. That result suggests a possible way to reduce query-time reading while retaining access to user-specific facts, but the abstract does not establish performance in deployed agents. The authors also study how different training recipes affect weight memory. They present it as a practical, low-cost component for persistent, multi-session agents, while identifying training and evaluation protocols for combining it with retrieval as an area their work informs. The evidence here is an abstract: it does not give benchmark details or establish how the reported results transfer beyond HaluMem.","question":"How does OPSD perform on factual-memory evaluations beyond HaluMem, compared with the text-memory baselines?","claims":[{"text":"The authors propose combining weight memory with text memory using a router that chooses which memory handles a query.","excerpt":"a verify-then-regenerate router that lets weight and text memory each cover the queries they handle best","support":"source-stated"},{"text":"The authors report that OPSD matches or surpasses strong text-memory baselines on HaluMem while reading fewer tokens per query.","excerpt":"On the HaluMem benchmark, OPSD matches or surpasses strong text-memory baselines, including full-context prompting and retrieval systems, while reading fewer tokens per query.","support":"source-stated"},{"text":"The authors present weight memory as a potential low-cost component for agents with persistent, multi-session use.","excerpt":"these results position weight memory as a practical, low-cost component of agent memory","support":"source-stated"},{"text":"The abstract describes a study of how training recipes affect weight memory's performance.","excerpt":"We further conduct an extensive study that characterizes the behavior of weight memory and analyzes how different training recipes affect its performance.","support":"source-stated"}],"limitations":["The available evidence is the arXiv abstract; it does not include benchmark details, experimental setup, or numerical results.","The abstract reports results on HaluMem but does not establish whether they transfer to other factual-memory evaluations or deployed agents."],"mediaId":"81e63e5f83366422d5ee782155ce52eac7594acaec64f2643ec765498e8a7e8b","correctionReason":null,"id":"3783528233dd2f0e437395c96b0985ca9ba91f235fd191349fc9ac3243e0f86a","publicationId":"92a3149b-8c77-4b6e-8dc6-8147b96c1b71","version":1,"sourceUrl":"https://arxiv.org/abs/2611.05636","sourceTitle":"Route to Remember: Adaptive Memory Selection for LLM Agents","publisher":"arXiv cs.AI","publishedAt":null,"originalPublished":{"value":"2026-11-03","precision":"day","basis":"publisher_citation_date"},"firstObservedAt":"2026-11-05T05:09:24.727Z","createdAt":"2026-11-09T11:20:09.449Z","historical":true,"entityIds":["arxiv"],"topicIds":["intelligence-and-discovery"],"media":[{"id":"81e63e5f83366422d5ee782155ce52eac7594acaec64f2643ec765498e8a7e8b","kind":"article","url":"https://arxiv.org/abs/2611.05636","sourceUrl":"https://arxiv.org/abs/2611.05636","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-11-09T11:19:11.102Z","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/2611.05636","checkedAt":"2026-11-09T11:52:10.428Z","items":[],"note":"No suitable media was found by the supported page extraction."},"editorialDecision":{"outcome":"accept","reasonCodes":["material_development"],"rubricVersion":"stream-admission-v5","mediaId":"81e63e5f83366422d5ee782155ce52eac7594acaec64f2643ec765498e8a7e8b","confidence":0.75,"provider":"typesafe","model":"jev-1.13.0","inputHash":"e1d5b213b80f9f4e6ee96e440b7b58ef906a024e6136d70597d10ef025ee13fb","decidedAt":"2026-11-09T11:19:14.325Z"},"provenance":{"textHash":"6a2b92cacde7286bb0fb7fd834651c7b3447e47db93073fd3fbfe0842a0d8260","observationId":"d8d480021b8280af6d720f32706366ccbe0c43bf5c39ff587fa9d8f160dc4a56","retrievedAt":"2026-11-09T11:19:10.495Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"531fd083-b8d8-492e-8a40-33d8c0e730a1"},"reviewer":{"model":"gpt-6-luna","runId":"51d722bb-6cf4-400f-8a47-9f78c809e199"}},"audit":{"status":"passed","reviewedAt":"2026-11-09T11:20:09.449Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Fang, Zeyu","url":null,"kind":"unknown"},{"name":"Yin, Jun","url":null,"kind":"unknown"},{"name":"Mei, Yongsheng","url":null,"kind":"unknown"},{"name":"Han, Xu","url":null,"kind":"unknown"},{"name":"Joshi, Chaitali","url":null,"kind":"unknown"},{"name":"Sun, Zhongkai","url":null,"kind":"unknown"},{"name":"Hao, Jie","url":null,"kind":"unknown"},{"name":"Guo, Bin","url":null,"kind":"unknown"},{"name":"Ma, Chengyuan","url":null,"kind":"unknown"},{"name":"Yao, Benjamin Z.","url":null,"kind":"unknown"},{"name":"Guo, Chenlei","url":null,"kind":"unknown"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Fang, Zeyu","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Fang, Zeyu","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.05636","locator":"head/meta[name=author|citation_author]"},"assertionId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T11:52:10.428Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52","sourceUrl":"https://arxiv.org/abs/2611.05636","contentSha256":"6527ecee471a2ee6df72ba0d8ef9157a7945c9fd3fe2704dd31c0ba49464a420","textHash":"6a2b92cacde7286bb0fb7fd834651c7b3447e47db93073fd3fbfe0842a0d8260","observedAt":"2026-11-09T11:52:10.428Z"}},{"actor":{"name":"Yin, Jun","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Yin, Jun","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.05636","locator":"head/meta[name=author|citation_author]"},"assertionId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52:participation:1","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T11:52:10.428Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52","sourceUrl":"https://arxiv.org/abs/2611.05636","contentSha256":"6527ecee471a2ee6df72ba0d8ef9157a7945c9fd3fe2704dd31c0ba49464a420","textHash":"6a2b92cacde7286bb0fb7fd834651c7b3447e47db93073fd3fbfe0842a0d8260","observedAt":"2026-11-09T11:52:10.428Z"}},{"actor":{"name":"Mei, Yongsheng","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Mei, Yongsheng","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.05636","locator":"head/meta[name=author|citation_author]"},"assertionId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52:participation:2","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T11:52:10.428Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52","sourceUrl":"https://arxiv.org/abs/2611.05636","contentSha256":"6527ecee471a2ee6df72ba0d8ef9157a7945c9fd3fe2704dd31c0ba49464a420","textHash":"6a2b92cacde7286bb0fb7fd834651c7b3447e47db93073fd3fbfe0842a0d8260","observedAt":"2026-11-09T11:52:10.428Z"}},{"actor":{"name":"Han, Xu","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Han, Xu","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.05636","locator":"head/meta[name=author|citation_author]"},"assertionId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52:participation:3","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T11:52:10.428Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52","sourceUrl":"https://arxiv.org/abs/2611.05636","contentSha256":"6527ecee471a2ee6df72ba0d8ef9157a7945c9fd3fe2704dd31c0ba49464a420","textHash":"6a2b92cacde7286bb0fb7fd834651c7b3447e47db93073fd3fbfe0842a0d8260","observedAt":"2026-11-09T11:52:10.428Z"}},{"actor":{"name":"Joshi, Chaitali","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Joshi, Chaitali","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.05636","locator":"head/meta[name=author|citation_author]"},"assertionId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52:participation:4","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T11:52:10.428Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52","sourceUrl":"https://arxiv.org/abs/2611.05636","contentSha256":"6527ecee471a2ee6df72ba0d8ef9157a7945c9fd3fe2704dd31c0ba49464a420","textHash":"6a2b92cacde7286bb0fb7fd834651c7b3447e47db93073fd3fbfe0842a0d8260","observedAt":"2026-11-09T11:52:10.428Z"}},{"actor":{"name":"Sun, Zhongkai","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Sun, Zhongkai","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.05636","locator":"head/meta[name=author|citation_author]"},"assertionId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52:participation:5","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T11:52:10.428Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52","sourceUrl":"https://arxiv.org/abs/2611.05636","contentSha256":"6527ecee471a2ee6df72ba0d8ef9157a7945c9fd3fe2704dd31c0ba49464a420","textHash":"6a2b92cacde7286bb0fb7fd834651c7b3447e47db93073fd3fbfe0842a0d8260","observedAt":"2026-11-09T11:52:10.428Z"}},{"actor":{"name":"Hao, Jie","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Hao, Jie","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.05636","locator":"head/meta[name=author|citation_author]"},"assertionId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52:participation:6","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T11:52:10.428Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52","sourceUrl":"https://arxiv.org/abs/2611.05636","contentSha256":"6527ecee471a2ee6df72ba0d8ef9157a7945c9fd3fe2704dd31c0ba49464a420","textHash":"6a2b92cacde7286bb0fb7fd834651c7b3447e47db93073fd3fbfe0842a0d8260","observedAt":"2026-11-09T11:52:10.428Z"}},{"actor":{"name":"Guo, Bin","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Guo, Bin","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.05636","locator":"head/meta[name=author|citation_author]"},"assertionId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52:participation:7","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T11:52:10.428Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52","sourceUrl":"https://arxiv.org/abs/2611.05636","contentSha256":"6527ecee471a2ee6df72ba0d8ef9157a7945c9fd3fe2704dd31c0ba49464a420","textHash":"6a2b92cacde7286bb0fb7fd834651c7b3447e47db93073fd3fbfe0842a0d8260","observedAt":"2026-11-09T11:52:10.428Z"}},{"actor":{"name":"Ma, Chengyuan","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Ma, Chengyuan","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.05636","locator":"head/meta[name=author|citation_author]"},"assertionId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52:participation:8","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T11:52:10.428Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52","sourceUrl":"https://arxiv.org/abs/2611.05636","contentSha256":"6527ecee471a2ee6df72ba0d8ef9157a7945c9fd3fe2704dd31c0ba49464a420","textHash":"6a2b92cacde7286bb0fb7fd834651c7b3447e47db93073fd3fbfe0842a0d8260","observedAt":"2026-11-09T11:52:10.428Z"}},{"actor":{"name":"Yao, Benjamin Z.","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Yao, Benjamin Z.","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.05636","locator":"head/meta[name=author|citation_author]"},"assertionId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52:participation:9","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T11:52:10.428Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52","sourceUrl":"https://arxiv.org/abs/2611.05636","contentSha256":"6527ecee471a2ee6df72ba0d8ef9157a7945c9fd3fe2704dd31c0ba49464a420","textHash":"6a2b92cacde7286bb0fb7fd834651c7b3447e47db93073fd3fbfe0842a0d8260","observedAt":"2026-11-09T11:52:10.428Z"}},{"actor":{"name":"Guo, Chenlei","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Guo, Chenlei","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.05636","locator":"head/meta[name=author|citation_author]"},"assertionId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52:participation:10","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T11:52:10.428Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52","sourceUrl":"https://arxiv.org/abs/2611.05636","contentSha256":"6527ecee471a2ee6df72ba0d8ef9157a7945c9fd3fe2704dd31c0ba49464a420","textHash":"6a2b92cacde7286bb0fb7fd834651c7b3447e47db93073fd3fbfe0842a0d8260","observedAt":"2026-11-09T11:52:10.428Z"}}],"participationCoverage":{"total":11,"returned":11,"complete":true,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":null,"sourceUrl":"https://arxiv.org/abs/2611.05636","snapshotId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52","contentSha256":"6527ecee471a2ee6df72ba0d8ef9157a7945c9fd3fe2704dd31c0ba49464a420","textHash":"6a2b92cacde7286bb0fb7fd834651c7b3447e47db93073fd3fbfe0842a0d8260","observedAt":"2026-11-09T11:52:10.428Z","basis":"publisher-page-metadata","sourceLinks":49,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/3783528233dd2f0e437395c96b0985ca9ba91f235fd191349fc9ac3243e0f86a/source"}},"canonicalUrl":"https://escapemeridian.com/readings/arxiv-route-to-remember-adaptive-memory-selection-for-llm-agents-37835282","formats":{"html":"/readings/arxiv-route-to-remember-adaptive-memory-selection-for-llm-agents-37835282","markdown":"/readings/arxiv-route-to-remember-adaptive-memory-selection-for-llm-agents-37835282.md","json":"/readings/arxiv-route-to-remember-adaptive-memory-selection-for-llm-agents-37835282.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"},{"id":"term-parameters","name":"Model parameters","explanation":"The internal numbers a model adjusts during training; their count is a rough measure of a model's size.","referenceUrl":"https://developers.google.com/machine-learning/glossary#parameter"},{"id":"term-benchmark","name":"Benchmark","explanation":"A defined set of tests used to compare systems. Results depend on the tasks and measures selected.","referenceUrl":"https://arxiv.org/abs/2211.09110"}],"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":"3783528233dd2f0e437395c96b0985ca9ba91f235fd191349fc9ac3243e0f86a","revision":"92a3149b-8c77-4b6e-8dc6-8147b96c1b71","revisionKind":"publication-id","title":"arXiv: Route to Remember: Adaptive Memory Selection for LLM Agents","url":"https://escapemeridian.com/readings/arxiv-route-to-remember-adaptive-memory-selection-for-llm-agents-37835282"},"source":{"type":"source","id":"https://arxiv.org/abs/2611.05636","revision":"6a2b92cacde7286bb0fb7fd834651c7b3447e47db93073fd3fbfe0842a0d8260","revisionKind":"sha256-source-text","title":"Route to Remember: Adaptive Memory Selection for LLM Agents","url":"https://arxiv.org/abs/2611.05636"},"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":"3783528233dd2f0e437395c96b0985ca9ba91f235fd191349fc9ac3243e0f86a","title":"arXiv: Route to Remember: Adaptive Memory Selection for LLM Agents","argument":"The authors of “Route to Remember: Adaptive Memory Selection for LLM Agents” argue that weight memory could help agents retain facts without adding text to every query. Weight memory stores knowledge in model parameters; the paper pairs it with text memory, which retrieves written information into the model’s context. Their method uses on-policy self-distillation (OPSD): a context-free student learns from a teacher model given the user’s memory. A verify-then-regenerate router selects between weight and text memory for each query. On the HaluMem benchmark, the authors report that OPSD matches or surpasses strong text-memory baselines, including full-context prompting and retrieval systems, while reading fewer tokens per query. That result suggests a possible way to reduce query-time reading while retaining access to user-specific facts, but the abstract does not establish performance in deployed agents. The authors also study how different training recipes affect weight memory. They present it as a practical, low-cost component for persistent, multi-session agents, while identifying training and evaluation protocols for combining it with retrieval as an area their work informs. The evidence here is an abstract: it does not give benchmark details or establish how the reported results transfer beyond HaluMem.","topicIds":["intelligence-and-discovery"],"domain":"science","claimType":"attributed reading","limits":"The available evidence is the arXiv abstract; it does not include benchmark details, experimental setup, or numerical results. The abstract reports results on HaluMem but does not establish whether they transfer to other factual-memory evaluations or deployed agents.","speaker":{"name":"Fang, Zeyu, Yin, Jun, Mei, Yongsheng, Han, Xu, Joshi, Chaitali, Sun, Zhongkai, Hao, Jie, Guo, Bin, Ma, Chengyuan, Yao, Benjamin Z., Guo, Chenlei","roleAtSource":"Article author"},"source":{"title":"Route to Remember: Adaptive Memory Selection for LLM Agents","publisher":"arXiv cs.AI","format":"Article","url":"https://arxiv.org/abs/2611.05636","publishedAt":"2026-11-03","dateBasis":"Date basis not provided."},"evidence":[{"text":"The authors propose combining weight memory with text memory using a router that chooses which memory handles a query.","excerpt":"a verify-then-regenerate router that lets weight and text memory each cover the queries they handle best","url":"https://arxiv.org/abs/2611.05636","relationship":"Source-stated"},{"text":"The authors report that OPSD matches or surpasses strong text-memory baselines on HaluMem while reading fewer tokens per query.","excerpt":"On the HaluMem benchmark, OPSD matches or surpasses strong text-memory baselines, including full-context prompting and retrieval systems, while reading fewer tokens per query.","url":"https://arxiv.org/abs/2611.05636","relationship":"Source-stated"},{"text":"The authors present weight memory as a potential low-cost component for agents with persistent, multi-session use.","excerpt":"these results position weight memory as a practical, low-cost component of agent memory","url":"https://arxiv.org/abs/2611.05636","relationship":"Source-stated"},{"text":"The abstract describes a study of how training recipes affect weight memory's performance.","excerpt":"We further conduct an extensive study that characterizes the behavior of weight memory and analyzes how different training recipes affect its performance.","url":"https://arxiv.org/abs/2611.05636","relationship":"Source-stated"}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2611.05636","checkedAt":"2026-11-09T11:52:10.428Z","items":[],"note":"No suitable media was found by the supported page extraction."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"How does OPSD perform on factual-memory evaluations beyond HaluMem, compared with the text-memory baselines?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"arXiv: Route to Remember: Adaptive Memory Selection for LLM Agents","summary":"The authors of “Route to Remember: Adaptive Memory Selection for LLM Agents” argue that weight memory could help agents retain facts without adding text to every query. Weight memory stores knowledge in model parameters; the paper pairs it with text memory, which retrieves written information into the model’s context. Their method uses on-policy self-distillation (OPSD): a context-free student learns from a teacher model given the user’s memory. A verify-then-regenerate router selects between weight and text memory for each query. On the HaluMem benchmark, the authors report that OPSD matches or surpasses strong text-memory baselines, including full-context prompting and retrieval systems, while reading fewer tokens per query. That result suggests a possible way to reduce query-time reading while retaining access to user-specific facts, but the abstract does not establish performance in deployed agents. The authors also study how different training recipes affect weight memory. They present it as a practical, low-cost component for persistent, multi-session agents, while identifying training and evaluation protocols for combining it with retrieval as an area their work informs. The evidence here is an abstract: it does not give benchmark details or establish how the reported results transfer beyond HaluMem.","question":"How does OPSD perform on factual-memory evaluations beyond HaluMem, compared with the text-memory baselines?","claims":[{"text":"The authors propose combining weight memory with text memory using a router that chooses which memory handles a query.","excerpt":"a verify-then-regenerate router that lets weight and text memory each cover the queries they handle best","support":"source-stated"},{"text":"The authors report that OPSD matches or surpasses strong text-memory baselines on HaluMem while reading fewer tokens per query.","excerpt":"On the HaluMem benchmark, OPSD matches or surpasses strong text-memory baselines, including full-context prompting and retrieval systems, while reading fewer tokens per query.","support":"source-stated"},{"text":"The authors present weight memory as a potential low-cost component for agents with persistent, multi-session use.","excerpt":"these results position weight memory as a practical, low-cost component of agent memory","support":"source-stated"},{"text":"The abstract describes a study of how training recipes affect weight memory's performance.","excerpt":"We further conduct an extensive study that characterizes the behavior of weight memory and analyzes how different training recipes affect its performance.","support":"source-stated"}],"limitations":["The available evidence is the arXiv abstract; it does not include benchmark details, experimental setup, or numerical results.","The abstract reports results on HaluMem but does not establish whether they transfer to other factual-memory evaluations or deployed agents."],"mediaId":"81e63e5f83366422d5ee782155ce52eac7594acaec64f2643ec765498e8a7e8b","correctionReason":null,"id":"3783528233dd2f0e437395c96b0985ca9ba91f235fd191349fc9ac3243e0f86a","publicationId":"92a3149b-8c77-4b6e-8dc6-8147b96c1b71","version":1,"sourceUrl":"https://arxiv.org/abs/2611.05636","sourceTitle":"Route to Remember: Adaptive Memory Selection for LLM Agents","publisher":"arXiv cs.AI","publishedAt":null,"originalPublished":{"value":"2026-11-03","precision":"day","basis":"publisher_citation_date"},"firstObservedAt":"2026-11-05T05:09:24.727Z","createdAt":"2026-11-09T11:20:09.449Z","historical":true,"entityIds":["arxiv"],"topicIds":["intelligence-and-discovery"],"media":[{"id":"81e63e5f83366422d5ee782155ce52eac7594acaec64f2643ec765498e8a7e8b","kind":"article","url":"https://arxiv.org/abs/2611.05636","sourceUrl":"https://arxiv.org/abs/2611.05636","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-11-09T11:19:11.102Z","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/2611.05636","checkedAt":"2026-11-09T11:52:10.428Z","items":[],"note":"No suitable media was found by the supported page extraction."},"editorialDecision":{"outcome":"accept","reasonCodes":["material_development"],"rubricVersion":"stream-admission-v5","mediaId":"81e63e5f83366422d5ee782155ce52eac7594acaec64f2643ec765498e8a7e8b","confidence":0.75,"provider":"typesafe","model":"jev-1.13.0","inputHash":"e1d5b213b80f9f4e6ee96e440b7b58ef906a024e6136d70597d10ef025ee13fb","decidedAt":"2026-11-09T11:19:14.325Z"},"provenance":{"textHash":"6a2b92cacde7286bb0fb7fd834651c7b3447e47db93073fd3fbfe0842a0d8260","observationId":"d8d480021b8280af6d720f32706366ccbe0c43bf5c39ff587fa9d8f160dc4a56","retrievedAt":"2026-11-09T11:19:10.495Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"531fd083-b8d8-492e-8a40-33d8c0e730a1"},"reviewer":{"model":"gpt-6-luna","runId":"51d722bb-6cf4-400f-8a47-9f78c809e199"}},"audit":{"status":"passed","reviewedAt":"2026-11-09T11:20:09.449Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Fang, Zeyu","url":null,"kind":"unknown"},{"name":"Yin, Jun","url":null,"kind":"unknown"},{"name":"Mei, Yongsheng","url":null,"kind":"unknown"},{"name":"Han, Xu","url":null,"kind":"unknown"},{"name":"Joshi, Chaitali","url":null,"kind":"unknown"},{"name":"Sun, Zhongkai","url":null,"kind":"unknown"},{"name":"Hao, Jie","url":null,"kind":"unknown"},{"name":"Guo, Bin","url":null,"kind":"unknown"},{"name":"Ma, Chengyuan","url":null,"kind":"unknown"},{"name":"Yao, Benjamin Z.","url":null,"kind":"unknown"},{"name":"Guo, Chenlei","url":null,"kind":"unknown"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Fang, Zeyu","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Fang, Zeyu","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.05636","locator":"head/meta[name=author|citation_author]"},"assertionId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T11:52:10.428Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52","sourceUrl":"https://arxiv.org/abs/2611.05636","contentSha256":"6527ecee471a2ee6df72ba0d8ef9157a7945c9fd3fe2704dd31c0ba49464a420","textHash":"6a2b92cacde7286bb0fb7fd834651c7b3447e47db93073fd3fbfe0842a0d8260","observedAt":"2026-11-09T11:52:10.428Z"}},{"actor":{"name":"Yin, Jun","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Yin, Jun","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.05636","locator":"head/meta[name=author|citation_author]"},"assertionId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52:participation:1","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T11:52:10.428Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52","sourceUrl":"https://arxiv.org/abs/2611.05636","contentSha256":"6527ecee471a2ee6df72ba0d8ef9157a7945c9fd3fe2704dd31c0ba49464a420","textHash":"6a2b92cacde7286bb0fb7fd834651c7b3447e47db93073fd3fbfe0842a0d8260","observedAt":"2026-11-09T11:52:10.428Z"}},{"actor":{"name":"Mei, Yongsheng","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Mei, Yongsheng","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.05636","locator":"head/meta[name=author|citation_author]"},"assertionId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52:participation:2","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T11:52:10.428Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52","sourceUrl":"https://arxiv.org/abs/2611.05636","contentSha256":"6527ecee471a2ee6df72ba0d8ef9157a7945c9fd3fe2704dd31c0ba49464a420","textHash":"6a2b92cacde7286bb0fb7fd834651c7b3447e47db93073fd3fbfe0842a0d8260","observedAt":"2026-11-09T11:52:10.428Z"}},{"actor":{"name":"Han, Xu","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Han, Xu","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.05636","locator":"head/meta[name=author|citation_author]"},"assertionId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52:participation:3","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T11:52:10.428Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52","sourceUrl":"https://arxiv.org/abs/2611.05636","contentSha256":"6527ecee471a2ee6df72ba0d8ef9157a7945c9fd3fe2704dd31c0ba49464a420","textHash":"6a2b92cacde7286bb0fb7fd834651c7b3447e47db93073fd3fbfe0842a0d8260","observedAt":"2026-11-09T11:52:10.428Z"}},{"actor":{"name":"Joshi, Chaitali","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Joshi, Chaitali","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.05636","locator":"head/meta[name=author|citation_author]"},"assertionId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52:participation:4","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T11:52:10.428Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52","sourceUrl":"https://arxiv.org/abs/2611.05636","contentSha256":"6527ecee471a2ee6df72ba0d8ef9157a7945c9fd3fe2704dd31c0ba49464a420","textHash":"6a2b92cacde7286bb0fb7fd834651c7b3447e47db93073fd3fbfe0842a0d8260","observedAt":"2026-11-09T11:52:10.428Z"}},{"actor":{"name":"Sun, Zhongkai","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Sun, Zhongkai","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.05636","locator":"head/meta[name=author|citation_author]"},"assertionId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52:participation:5","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T11:52:10.428Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52","sourceUrl":"https://arxiv.org/abs/2611.05636","contentSha256":"6527ecee471a2ee6df72ba0d8ef9157a7945c9fd3fe2704dd31c0ba49464a420","textHash":"6a2b92cacde7286bb0fb7fd834651c7b3447e47db93073fd3fbfe0842a0d8260","observedAt":"2026-11-09T11:52:10.428Z"}},{"actor":{"name":"Hao, Jie","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Hao, Jie","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.05636","locator":"head/meta[name=author|citation_author]"},"assertionId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52:participation:6","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T11:52:10.428Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52","sourceUrl":"https://arxiv.org/abs/2611.05636","contentSha256":"6527ecee471a2ee6df72ba0d8ef9157a7945c9fd3fe2704dd31c0ba49464a420","textHash":"6a2b92cacde7286bb0fb7fd834651c7b3447e47db93073fd3fbfe0842a0d8260","observedAt":"2026-11-09T11:52:10.428Z"}},{"actor":{"name":"Guo, Bin","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Guo, Bin","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.05636","locator":"head/meta[name=author|citation_author]"},"assertionId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52:participation:7","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T11:52:10.428Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52","sourceUrl":"https://arxiv.org/abs/2611.05636","contentSha256":"6527ecee471a2ee6df72ba0d8ef9157a7945c9fd3fe2704dd31c0ba49464a420","textHash":"6a2b92cacde7286bb0fb7fd834651c7b3447e47db93073fd3fbfe0842a0d8260","observedAt":"2026-11-09T11:52:10.428Z"}},{"actor":{"name":"Ma, Chengyuan","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Ma, Chengyuan","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.05636","locator":"head/meta[name=author|citation_author]"},"assertionId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52:participation:8","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T11:52:10.428Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52","sourceUrl":"https://arxiv.org/abs/2611.05636","contentSha256":"6527ecee471a2ee6df72ba0d8ef9157a7945c9fd3fe2704dd31c0ba49464a420","textHash":"6a2b92cacde7286bb0fb7fd834651c7b3447e47db93073fd3fbfe0842a0d8260","observedAt":"2026-11-09T11:52:10.428Z"}},{"actor":{"name":"Yao, Benjamin Z.","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Yao, Benjamin Z.","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.05636","locator":"head/meta[name=author|citation_author]"},"assertionId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52:participation:9","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T11:52:10.428Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52","sourceUrl":"https://arxiv.org/abs/2611.05636","contentSha256":"6527ecee471a2ee6df72ba0d8ef9157a7945c9fd3fe2704dd31c0ba49464a420","textHash":"6a2b92cacde7286bb0fb7fd834651c7b3447e47db93073fd3fbfe0842a0d8260","observedAt":"2026-11-09T11:52:10.428Z"}},{"actor":{"name":"Guo, Chenlei","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Guo, Chenlei","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.05636","locator":"head/meta[name=author|citation_author]"},"assertionId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52:participation:10","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T11:52:10.428Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52","sourceUrl":"https://arxiv.org/abs/2611.05636","contentSha256":"6527ecee471a2ee6df72ba0d8ef9157a7945c9fd3fe2704dd31c0ba49464a420","textHash":"6a2b92cacde7286bb0fb7fd834651c7b3447e47db93073fd3fbfe0842a0d8260","observedAt":"2026-11-09T11:52:10.428Z"}}],"participationCoverage":{"total":11,"returned":11,"complete":true,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":null,"sourceUrl":"https://arxiv.org/abs/2611.05636","snapshotId":"74a921effc97b1ab228464af34e660194cace8ffc8ddc6f36ea3d17563ff4f52","contentSha256":"6527ecee471a2ee6df72ba0d8ef9157a7945c9fd3fe2704dd31c0ba49464a420","textHash":"6a2b92cacde7286bb0fb7fd834651c7b3447e47db93073fd3fbfe0842a0d8260","observedAt":"2026-11-09T11:52:10.428Z","basis":"publisher-page-metadata","sourceLinks":49,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/3783528233dd2f0e437395c96b0985ca9ba91f235fd191349fc9ac3243e0f86a/source"}}}],"audits":[{"publicationId":"92a3149b-8c77-4b6e-8dc6-8147b96c1b71","verdict":"passed","findings":[{"dimension":"attribution","status":"pass","explanation":"The reading consistently attributes the method, results, and framing to the authors of the arXiv paper, using phrases such as 'the authors report' and 'they present it'. Source assertions are not presented as established fact.","excerpt":"Authors:Zeyu Fang, Jun Yin, Yongsheng Mei, Xu Han, Chaitali Joshi, Zhongkai Sun, Jie Hao, Bin Guo, Chengyuan Ma, Benjamin Z. Yao, Chenlei Guo"},{"dimension":"summary","status":"pass","explanation":"The reading accurately explains weight memory, text memory, OPSD, the router, and the HaluMem result in accessible terms. One optional repair: the teacher is the same model conditioned on the user's memory, which is the 'self' in self-distillation; adding 'the same model' would sharpen this. The cost-growth motivation is only implied by 'without adding text to every query'. Neither gap prevents understanding. The question about results beyond HaluMem follows from the evidence.","excerpt":"It trains a context-free student from a teacher, which is the same model conditioned on the user's memory"},{"dimension":"date","status":"pass","explanation":"originalPublished is 2026-11-03, which matches the arXiv v1 submission date. The feed date of 2026-11-05 is later and correctly not used.","excerpt":"[Submitted on 3 Nov 2026]"},{"dimension":"support","status":"pass","explanation":"Each claim excerpt is verbatim and supports its claim. The inference about reducing query-time reading is explicitly conditional and is paired with the caveat that deployed performance is not established. The limitations correctly note that the abstract lacks numerical and setup details. No benefits or costs beyond the source are invented.","excerpt":"On the HaluMem benchmark, OPSD matches or surpasses strong text-memory baselines, including full-context prompting and retrieval systems, while reading fewer tokens per query."},{"dimension":"media","status":"pass","explanation":"The selected media is the verified direct arXiv abstract page for this paper, which is the original source.","excerpt":null}],"createdAt":"2026-11-09T11:20:09.449Z","model":"anthropic/claude-opus-5.5"}],"correctionUrl":"https://escapemeridian.com/corrections/"}