{"id":"ae542be642e84d180db0ae28936bd1514d68739472fe654d8675093d6d1c5d79","title":"Axis-LoRA targets time, place, person and event in personal-memory retrieval","argument":"Authors Dohoon Kim, Junho Park, Hyunwoong Bae, Seyoon Kim and Taesup Moon argue that frozen dense embedding models used in personal-memory retrieval do not distinguish memories well by time, place, person and event. They report that changing a memory’s date or place barely shifts its embedding, and that a trained projection on top of the frozen embedding does not recover the missing signal. Their explanation is that each axis is encoded by a distinct attention-head pattern inside the model. They propose Axis-LoRA, which places four LoRA branches inside the embedding model and combines them with a per-query gate. The authors evaluate Axis-LoRA across three standard embedding models on LongMemEval, ATM-Bench and LoCoMo. They report a 4-benchmark mean R@10 improvement of +1.28 to +1.56 pp, while training only a small fraction of the base parameters on a small axis-counterfactual training set. The work’s significance, as presented, is a targeted way to adapt retrieval for personal memories without training most of the base model. The abstract reports benchmark results; how those results translate to personal-memory retrieval in use remains an open question.","topicIds":["intelligence-and-discovery"],"domain":"science","claimType":"attributed reading","limits":"The supplied evidence is the paper’s abstract, not its full methods or results; it does not provide the detailed evaluation setup or benchmark-by-benchmark results. The reported improvements are benchmark results; the supplied evidence does not establish how Axis-LoRA performs in personal-memory retrieval in use.","speaker":{"name":"Kim, Dohoon, Park, Junho, Bae, Hyunwoong, Kim, Seyoon, Moon, Taesup","roleAtSource":"Article author"},"source":{"title":"Axis-LoRA: Probing and Adapting Embedding Models for Personal-Memory Retrieval","publisher":"arXiv cs.AI","format":"Article","url":"https://arxiv.org/abs/2611.07820","publishedAt":"2026-11-05","dateBasis":"Date basis not provided."},"evidence":[{"text":"The authors say frozen dense embedding models do not capture time, place, person and event well for personal-memory retrieval.","excerpt":"we find that these models do not capture the four axes well","url":"https://arxiv.org/abs/2611.07820","relationship":"Source-stated"},{"text":"The authors report that changing a memory’s date or place barely shifts its embedding, and a trained projection on the frozen embedding does not recover the missing signal.","excerpt":"Swapping a memory's date or place barely shifts its embedding, and a trained projection on top of the frozen embedding does not recover the missing signal.","url":"https://arxiv.org/abs/2611.07820","relationship":"Source-stated"},{"text":"The authors propose Axis-LoRA, using four LoRA branches inside the embedding model and a per-query gate.","excerpt":"we propose Axis-LoRA, a per-axis adapter that places four LoRA branches inside the embedding model and combines them via a per-query gate.","url":"https://arxiv.org/abs/2611.07820","relationship":"Source-stated"},{"text":"The authors report a 4-benchmark mean R@10 improvement of +1.28 to +1.56 pp across three standard embedding models evaluated on LongMemEval, ATM-Bench and LoCoMo.","excerpt":"Across three standard embedding models evaluated on LongMemEval, ATM-Bench, and LoCoMo, Axis-LoRA improves 4-benchmark mean R@10 by $+1.28$ to $+1.56$~pp","url":"https://arxiv.org/abs/2611.07820","relationship":"Source-stated"}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2611.07820","checkedAt":"2026-11-09T04:56:08.987Z","items":[],"note":"No suitable media was found by the supported page extraction."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"How does Axis-LoRA perform on personal-memory retrieval beyond the LongMemEval, ATM-Bench and LoCoMo evaluations?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"Axis-LoRA targets time, place, person and event in personal-memory retrieval","summary":"Authors Dohoon Kim, Junho Park, Hyunwoong Bae, Seyoon Kim and Taesup Moon argue that frozen dense embedding models used in personal-memory retrieval do not distinguish memories well by time, place, person and event. They report that changing a memory’s date or place barely shifts its embedding, and that a trained projection on top of the frozen embedding does not recover the missing signal. Their explanation is that each axis is encoded by a distinct attention-head pattern inside the model. They propose Axis-LoRA, which places four LoRA branches inside the embedding model and combines them with a per-query gate. The authors evaluate Axis-LoRA across three standard embedding models on LongMemEval, ATM-Bench and LoCoMo. They report a 4-benchmark mean R@10 improvement of +1.28 to +1.56 pp, while training only a small fraction of the base parameters on a small axis-counterfactual training set. The work’s significance, as presented, is a targeted way to adapt retrieval for personal memories without training most of the base model. The abstract reports benchmark results; how those results translate to personal-memory retrieval in use remains an open question.","question":"How does Axis-LoRA perform on personal-memory retrieval beyond the LongMemEval, ATM-Bench and LoCoMo evaluations?","claims":[{"text":"The authors say frozen dense embedding models do not capture time, place, person and event well for personal-memory retrieval.","excerpt":"we find that these models do not capture the four axes well","support":"source-stated"},{"text":"The authors report that changing a memory’s date or place barely shifts its embedding, and a trained projection on the frozen embedding does not recover the missing signal.","excerpt":"Swapping a memory's date or place barely shifts its embedding, and a trained projection on top of the frozen embedding does not recover the missing signal.","support":"source-stated"},{"text":"The authors propose Axis-LoRA, using four LoRA branches inside the embedding model and a per-query gate.","excerpt":"we propose Axis-LoRA, a per-axis adapter that places four LoRA branches inside the embedding model and combines them via a per-query gate.","support":"source-stated"},{"text":"The authors report a 4-benchmark mean R@10 improvement of +1.28 to +1.56 pp across three standard embedding models evaluated on LongMemEval, ATM-Bench and LoCoMo.","excerpt":"Across three standard embedding models evaluated on LongMemEval, ATM-Bench, and LoCoMo, Axis-LoRA improves 4-benchmark mean R@10 by $+1.28$ to $+1.56$~pp","support":"source-stated"}],"limitations":["The supplied evidence is the paper’s abstract, not its full methods or results; it does not provide the detailed evaluation setup or benchmark-by-benchmark results.","The reported improvements are benchmark results; the supplied evidence does not establish how Axis-LoRA performs in personal-memory retrieval in use."],"mediaId":"90625fa0d9224b8680fba79324726a86c7875698ef2041218779b5b2a2a50020","correctionReason":null,"id":"ae542be642e84d180db0ae28936bd1514d68739472fe654d8675093d6d1c5d79","publicationId":"5717f61a-94b5-447d-8aa0-b931066b6163","version":1,"sourceUrl":"https://arxiv.org/abs/2611.07820","sourceTitle":"Axis-LoRA: Probing and Adapting Embedding Models for Personal-Memory Retrieval","publisher":"arXiv cs.AI","publishedAt":null,"originalPublished":{"value":"2026-11-05","precision":"day","basis":"publisher_citation_date"},"firstObservedAt":"2026-11-06T05:10:58.261Z","createdAt":"2026-11-09T03:53:10.211Z","historical":true,"entityIds":["arxiv"],"topicIds":["intelligence-and-discovery"],"media":[{"id":"90625fa0d9224b8680fba79324726a86c7875698ef2041218779b5b2a2a50020","kind":"article","url":"https://arxiv.org/abs/2611.07820","sourceUrl":"https://arxiv.org/abs/2611.07820","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-11-09T03:40:16.536Z","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/2611.07820","checkedAt":"2026-11-09T04:56:08.987Z","items":[],"note":"No suitable media was found by the supported page extraction."},"editorialDecision":{"outcome":"accept","reasonCodes":["material_development"],"rubricVersion":"stream-admission-v5","mediaId":"90625fa0d9224b8680fba79324726a86c7875698ef2041218779b5b2a2a50020","confidence":0.83,"provider":"typesafe","model":"jev-1.13.0","inputHash":"e7dceeb46479a6960a0c377069526b253a81d1a7e9f5309afc1dedd3fd579ecb","decidedAt":"2026-11-09T03:40:17.186Z"},"provenance":{"textHash":"42f8da428d132245a9f8f7b672d0e61e530881d3b12c36ab9209b1cf6be640ce","observationId":"ca2b174047972c1d5bced6158e9aa534c9ee5d48f3b95ca7685f353d8eab6771","retrievedAt":"2026-11-09T03:40:16.420Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"64b767ee-81e0-4c87-8e09-89d949f373c0"},"reviewer":{"model":"gpt-6-luna","runId":"815b4124-9067-44dc-829b-338366ecc513"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-09T03:53:10.211Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Kim, Dohoon","url":null,"kind":"unknown"},{"name":"Park, Junho","url":null,"kind":"unknown"},{"name":"Bae, Hyunwoong","url":null,"kind":"unknown"},{"name":"Kim, Seyoon","url":null,"kind":"unknown"},{"name":"Moon, Taesup","url":null,"kind":"unknown"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Kim, Dohoon","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Kim, Dohoon","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.07820","locator":"head/meta[name=author|citation_author]"},"assertionId":"1b660b59fa7bb6fdb305844b91e8cbe5f39ce1ba9480f7c9a439f43670031196:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T04:56:08.987Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"1b660b59fa7bb6fdb305844b91e8cbe5f39ce1ba9480f7c9a439f43670031196","sourceUrl":"https://arxiv.org/abs/2611.07820","contentSha256":"f16419606dbd19b0822c24ac38473df33737086ab9120f140b898e77d7688899","textHash":"42f8da428d132245a9f8f7b672d0e61e530881d3b12c36ab9209b1cf6be640ce","observedAt":"2026-11-09T04:56:08.987Z"}},{"actor":{"name":"Park, Junho","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Park, Junho","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.07820","locator":"head/meta[name=author|citation_author]"},"assertionId":"1b660b59fa7bb6fdb305844b91e8cbe5f39ce1ba9480f7c9a439f43670031196:participation:1","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T04:56:08.987Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"1b660b59fa7bb6fdb305844b91e8cbe5f39ce1ba9480f7c9a439f43670031196","sourceUrl":"https://arxiv.org/abs/2611.07820","contentSha256":"f16419606dbd19b0822c24ac38473df33737086ab9120f140b898e77d7688899","textHash":"42f8da428d132245a9f8f7b672d0e61e530881d3b12c36ab9209b1cf6be640ce","observedAt":"2026-11-09T04:56:08.987Z"}},{"actor":{"name":"Bae, Hyunwoong","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Bae, Hyunwoong","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.07820","locator":"head/meta[name=author|citation_author]"},"assertionId":"1b660b59fa7bb6fdb305844b91e8cbe5f39ce1ba9480f7c9a439f43670031196:participation:2","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T04:56:08.987Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"1b660b59fa7bb6fdb305844b91e8cbe5f39ce1ba9480f7c9a439f43670031196","sourceUrl":"https://arxiv.org/abs/2611.07820","contentSha256":"f16419606dbd19b0822c24ac38473df33737086ab9120f140b898e77d7688899","textHash":"42f8da428d132245a9f8f7b672d0e61e530881d3b12c36ab9209b1cf6be640ce","observedAt":"2026-11-09T04:56:08.987Z"}},{"actor":{"name":"Kim, Seyoon","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Kim, Seyoon","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.07820","locator":"head/meta[name=author|citation_author]"},"assertionId":"1b660b59fa7bb6fdb305844b91e8cbe5f39ce1ba9480f7c9a439f43670031196:participation:3","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T04:56:08.987Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"1b660b59fa7bb6fdb305844b91e8cbe5f39ce1ba9480f7c9a439f43670031196","sourceUrl":"https://arxiv.org/abs/2611.07820","contentSha256":"f16419606dbd19b0822c24ac38473df33737086ab9120f140b898e77d7688899","textHash":"42f8da428d132245a9f8f7b672d0e61e530881d3b12c36ab9209b1cf6be640ce","observedAt":"2026-11-09T04:56:08.987Z"}},{"actor":{"name":"Moon, Taesup","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Moon, Taesup","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.07820","locator":"head/meta[name=author|citation_author]"},"assertionId":"1b660b59fa7bb6fdb305844b91e8cbe5f39ce1ba9480f7c9a439f43670031196:participation:4","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T04:56:08.987Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"1b660b59fa7bb6fdb305844b91e8cbe5f39ce1ba9480f7c9a439f43670031196","sourceUrl":"https://arxiv.org/abs/2611.07820","contentSha256":"f16419606dbd19b0822c24ac38473df33737086ab9120f140b898e77d7688899","textHash":"42f8da428d132245a9f8f7b672d0e61e530881d3b12c36ab9209b1cf6be640ce","observedAt":"2026-11-09T04:56:08.987Z"}}],"participationCoverage":{"total":5,"returned":5,"complete":true,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":null,"sourceUrl":"https://arxiv.org/abs/2611.07820","snapshotId":"1b660b59fa7bb6fdb305844b91e8cbe5f39ce1ba9480f7c9a439f43670031196","contentSha256":"f16419606dbd19b0822c24ac38473df33737086ab9120f140b898e77d7688899","textHash":"42f8da428d132245a9f8f7b672d0e61e530881d3b12c36ab9209b1cf6be640ce","observedAt":"2026-11-09T04:56:08.987Z","basis":"publisher-page-metadata","sourceLinks":45,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/ae542be642e84d180db0ae28936bd1514d68739472fe654d8675093d6d1c5d79/source"}},"canonicalUrl":"https://escapemeridian.com/readings/axis-lora-targets-time-place-person-and-event-in-personal-ae542be6","formats":{"html":"/readings/axis-lora-targets-time-place-person-and-event-in-personal-ae542be6","markdown":"/readings/axis-lora-targets-time-place-person-and-event-in-personal-ae542be6.md","json":"/readings/axis-lora-targets-time-place-person-and-event-in-personal-ae542be6.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-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"},{"id":"term-counterfactual","name":"Counterfactual","explanation":"An argument or scenario that considers what would have resulted if events had gone differently from how they actually did, such as if a policy had never been adopted; as an adjective, relating to this kind of thinking about what did not happen but could have.","referenceUrl":"https://dictionary.cambridge.org/dictionary/english/counterfactual"}],"entities":[{"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"},{"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"}]},"researchQuestions":[],"withdrawn":false,"references":{"schemaVersion":1,"target":{"type":"reading","id":"ae542be642e84d180db0ae28936bd1514d68739472fe654d8675093d6d1c5d79","revision":"5717f61a-94b5-447d-8aa0-b931066b6163","revisionKind":"publication-id","title":"Axis-LoRA targets time, place, person and event in personal-memory retrieval","url":"https://escapemeridian.com/readings/axis-lora-targets-time-place-person-and-event-in-personal-ae542be6"},"source":{"type":"source","id":"https://arxiv.org/abs/2611.07820","revision":"42f8da428d132245a9f8f7b672d0e61e530881d3b12c36ab9209b1cf6be640ce","revisionKind":"sha256-source-text","title":"Axis-LoRA: Probing and Adapting Embedding Models for Personal-Memory Retrieval","url":"https://arxiv.org/abs/2611.07820"},"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":"ae542be642e84d180db0ae28936bd1514d68739472fe654d8675093d6d1c5d79","title":"Axis-LoRA targets time, place, person and event in personal-memory retrieval","argument":"Authors Dohoon Kim, Junho Park, Hyunwoong Bae, Seyoon Kim and Taesup Moon argue that frozen dense embedding models used in personal-memory retrieval do not distinguish memories well by time, place, person and event. They report that changing a memory’s date or place barely shifts its embedding, and that a trained projection on top of the frozen embedding does not recover the missing signal. Their explanation is that each axis is encoded by a distinct attention-head pattern inside the model. They propose Axis-LoRA, which places four LoRA branches inside the embedding model and combines them with a per-query gate. The authors evaluate Axis-LoRA across three standard embedding models on LongMemEval, ATM-Bench and LoCoMo. They report a 4-benchmark mean R@10 improvement of +1.28 to +1.56 pp, while training only a small fraction of the base parameters on a small axis-counterfactual training set. The work’s significance, as presented, is a targeted way to adapt retrieval for personal memories without training most of the base model. The abstract reports benchmark results; how those results translate to personal-memory retrieval in use remains an open question.","topicIds":["intelligence-and-discovery"],"domain":"science","claimType":"attributed reading","limits":"The supplied evidence is the paper’s abstract, not its full methods or results; it does not provide the detailed evaluation setup or benchmark-by-benchmark results. The reported improvements are benchmark results; the supplied evidence does not establish how Axis-LoRA performs in personal-memory retrieval in use.","speaker":{"name":"Kim, Dohoon, Park, Junho, Bae, Hyunwoong, Kim, Seyoon, Moon, Taesup","roleAtSource":"Article author"},"source":{"title":"Axis-LoRA: Probing and Adapting Embedding Models for Personal-Memory Retrieval","publisher":"arXiv cs.AI","format":"Article","url":"https://arxiv.org/abs/2611.07820","publishedAt":"2026-11-05","dateBasis":"Date basis not provided."},"evidence":[{"text":"The authors say frozen dense embedding models do not capture time, place, person and event well for personal-memory retrieval.","excerpt":"we find that these models do not capture the four axes well","url":"https://arxiv.org/abs/2611.07820","relationship":"Source-stated"},{"text":"The authors report that changing a memory’s date or place barely shifts its embedding, and a trained projection on the frozen embedding does not recover the missing signal.","excerpt":"Swapping a memory's date or place barely shifts its embedding, and a trained projection on top of the frozen embedding does not recover the missing signal.","url":"https://arxiv.org/abs/2611.07820","relationship":"Source-stated"},{"text":"The authors propose Axis-LoRA, using four LoRA branches inside the embedding model and a per-query gate.","excerpt":"we propose Axis-LoRA, a per-axis adapter that places four LoRA branches inside the embedding model and combines them via a per-query gate.","url":"https://arxiv.org/abs/2611.07820","relationship":"Source-stated"},{"text":"The authors report a 4-benchmark mean R@10 improvement of +1.28 to +1.56 pp across three standard embedding models evaluated on LongMemEval, ATM-Bench and LoCoMo.","excerpt":"Across three standard embedding models evaluated on LongMemEval, ATM-Bench, and LoCoMo, Axis-LoRA improves 4-benchmark mean R@10 by $+1.28$ to $+1.56$~pp","url":"https://arxiv.org/abs/2611.07820","relationship":"Source-stated"}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2611.07820","checkedAt":"2026-11-09T04:56:08.987Z","items":[],"note":"No suitable media was found by the supported page extraction."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"How does Axis-LoRA perform on personal-memory retrieval beyond the LongMemEval, ATM-Bench and LoCoMo evaluations?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"Axis-LoRA targets time, place, person and event in personal-memory retrieval","summary":"Authors Dohoon Kim, Junho Park, Hyunwoong Bae, Seyoon Kim and Taesup Moon argue that frozen dense embedding models used in personal-memory retrieval do not distinguish memories well by time, place, person and event. They report that changing a memory’s date or place barely shifts its embedding, and that a trained projection on top of the frozen embedding does not recover the missing signal. Their explanation is that each axis is encoded by a distinct attention-head pattern inside the model. They propose Axis-LoRA, which places four LoRA branches inside the embedding model and combines them with a per-query gate. The authors evaluate Axis-LoRA across three standard embedding models on LongMemEval, ATM-Bench and LoCoMo. They report a 4-benchmark mean R@10 improvement of +1.28 to +1.56 pp, while training only a small fraction of the base parameters on a small axis-counterfactual training set. The work’s significance, as presented, is a targeted way to adapt retrieval for personal memories without training most of the base model. The abstract reports benchmark results; how those results translate to personal-memory retrieval in use remains an open question.","question":"How does Axis-LoRA perform on personal-memory retrieval beyond the LongMemEval, ATM-Bench and LoCoMo evaluations?","claims":[{"text":"The authors say frozen dense embedding models do not capture time, place, person and event well for personal-memory retrieval.","excerpt":"we find that these models do not capture the four axes well","support":"source-stated"},{"text":"The authors report that changing a memory’s date or place barely shifts its embedding, and a trained projection on the frozen embedding does not recover the missing signal.","excerpt":"Swapping a memory's date or place barely shifts its embedding, and a trained projection on top of the frozen embedding does not recover the missing signal.","support":"source-stated"},{"text":"The authors propose Axis-LoRA, using four LoRA branches inside the embedding model and a per-query gate.","excerpt":"we propose Axis-LoRA, a per-axis adapter that places four LoRA branches inside the embedding model and combines them via a per-query gate.","support":"source-stated"},{"text":"The authors report a 4-benchmark mean R@10 improvement of +1.28 to +1.56 pp across three standard embedding models evaluated on LongMemEval, ATM-Bench and LoCoMo.","excerpt":"Across three standard embedding models evaluated on LongMemEval, ATM-Bench, and LoCoMo, Axis-LoRA improves 4-benchmark mean R@10 by $+1.28$ to $+1.56$~pp","support":"source-stated"}],"limitations":["The supplied evidence is the paper’s abstract, not its full methods or results; it does not provide the detailed evaluation setup or benchmark-by-benchmark results.","The reported improvements are benchmark results; the supplied evidence does not establish how Axis-LoRA performs in personal-memory retrieval in use."],"mediaId":"90625fa0d9224b8680fba79324726a86c7875698ef2041218779b5b2a2a50020","correctionReason":null,"id":"ae542be642e84d180db0ae28936bd1514d68739472fe654d8675093d6d1c5d79","publicationId":"5717f61a-94b5-447d-8aa0-b931066b6163","version":1,"sourceUrl":"https://arxiv.org/abs/2611.07820","sourceTitle":"Axis-LoRA: Probing and Adapting Embedding Models for Personal-Memory Retrieval","publisher":"arXiv cs.AI","publishedAt":null,"originalPublished":{"value":"2026-11-05","precision":"day","basis":"publisher_citation_date"},"firstObservedAt":"2026-11-06T05:10:58.261Z","createdAt":"2026-11-09T03:53:10.211Z","historical":true,"entityIds":["arxiv"],"topicIds":["intelligence-and-discovery"],"media":[{"id":"90625fa0d9224b8680fba79324726a86c7875698ef2041218779b5b2a2a50020","kind":"article","url":"https://arxiv.org/abs/2611.07820","sourceUrl":"https://arxiv.org/abs/2611.07820","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-11-09T03:40:16.536Z","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/2611.07820","checkedAt":"2026-11-09T04:56:08.987Z","items":[],"note":"No suitable media was found by the supported page extraction."},"editorialDecision":{"outcome":"accept","reasonCodes":["material_development"],"rubricVersion":"stream-admission-v5","mediaId":"90625fa0d9224b8680fba79324726a86c7875698ef2041218779b5b2a2a50020","confidence":0.83,"provider":"typesafe","model":"jev-1.13.0","inputHash":"e7dceeb46479a6960a0c377069526b253a81d1a7e9f5309afc1dedd3fd579ecb","decidedAt":"2026-11-09T03:40:17.186Z"},"provenance":{"textHash":"42f8da428d132245a9f8f7b672d0e61e530881d3b12c36ab9209b1cf6be640ce","observationId":"ca2b174047972c1d5bced6158e9aa534c9ee5d48f3b95ca7685f353d8eab6771","retrievedAt":"2026-11-09T03:40:16.420Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"64b767ee-81e0-4c87-8e09-89d949f373c0"},"reviewer":{"model":"gpt-6-luna","runId":"815b4124-9067-44dc-829b-338366ecc513"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-09T03:53:10.211Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Kim, Dohoon","url":null,"kind":"unknown"},{"name":"Park, Junho","url":null,"kind":"unknown"},{"name":"Bae, Hyunwoong","url":null,"kind":"unknown"},{"name":"Kim, Seyoon","url":null,"kind":"unknown"},{"name":"Moon, Taesup","url":null,"kind":"unknown"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Kim, Dohoon","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Kim, Dohoon","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.07820","locator":"head/meta[name=author|citation_author]"},"assertionId":"1b660b59fa7bb6fdb305844b91e8cbe5f39ce1ba9480f7c9a439f43670031196:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T04:56:08.987Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"1b660b59fa7bb6fdb305844b91e8cbe5f39ce1ba9480f7c9a439f43670031196","sourceUrl":"https://arxiv.org/abs/2611.07820","contentSha256":"f16419606dbd19b0822c24ac38473df33737086ab9120f140b898e77d7688899","textHash":"42f8da428d132245a9f8f7b672d0e61e530881d3b12c36ab9209b1cf6be640ce","observedAt":"2026-11-09T04:56:08.987Z"}},{"actor":{"name":"Park, Junho","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Park, Junho","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.07820","locator":"head/meta[name=author|citation_author]"},"assertionId":"1b660b59fa7bb6fdb305844b91e8cbe5f39ce1ba9480f7c9a439f43670031196:participation:1","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T04:56:08.987Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"1b660b59fa7bb6fdb305844b91e8cbe5f39ce1ba9480f7c9a439f43670031196","sourceUrl":"https://arxiv.org/abs/2611.07820","contentSha256":"f16419606dbd19b0822c24ac38473df33737086ab9120f140b898e77d7688899","textHash":"42f8da428d132245a9f8f7b672d0e61e530881d3b12c36ab9209b1cf6be640ce","observedAt":"2026-11-09T04:56:08.987Z"}},{"actor":{"name":"Bae, Hyunwoong","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Bae, Hyunwoong","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.07820","locator":"head/meta[name=author|citation_author]"},"assertionId":"1b660b59fa7bb6fdb305844b91e8cbe5f39ce1ba9480f7c9a439f43670031196:participation:2","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T04:56:08.987Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"1b660b59fa7bb6fdb305844b91e8cbe5f39ce1ba9480f7c9a439f43670031196","sourceUrl":"https://arxiv.org/abs/2611.07820","contentSha256":"f16419606dbd19b0822c24ac38473df33737086ab9120f140b898e77d7688899","textHash":"42f8da428d132245a9f8f7b672d0e61e530881d3b12c36ab9209b1cf6be640ce","observedAt":"2026-11-09T04:56:08.987Z"}},{"actor":{"name":"Kim, Seyoon","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Kim, Seyoon","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.07820","locator":"head/meta[name=author|citation_author]"},"assertionId":"1b660b59fa7bb6fdb305844b91e8cbe5f39ce1ba9480f7c9a439f43670031196:participation:3","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T04:56:08.987Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"1b660b59fa7bb6fdb305844b91e8cbe5f39ce1ba9480f7c9a439f43670031196","sourceUrl":"https://arxiv.org/abs/2611.07820","contentSha256":"f16419606dbd19b0822c24ac38473df33737086ab9120f140b898e77d7688899","textHash":"42f8da428d132245a9f8f7b672d0e61e530881d3b12c36ab9209b1cf6be640ce","observedAt":"2026-11-09T04:56:08.987Z"}},{"actor":{"name":"Moon, Taesup","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Moon, Taesup","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.07820","locator":"head/meta[name=author|citation_author]"},"assertionId":"1b660b59fa7bb6fdb305844b91e8cbe5f39ce1ba9480f7c9a439f43670031196:participation:4","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T04:56:08.987Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"1b660b59fa7bb6fdb305844b91e8cbe5f39ce1ba9480f7c9a439f43670031196","sourceUrl":"https://arxiv.org/abs/2611.07820","contentSha256":"f16419606dbd19b0822c24ac38473df33737086ab9120f140b898e77d7688899","textHash":"42f8da428d132245a9f8f7b672d0e61e530881d3b12c36ab9209b1cf6be640ce","observedAt":"2026-11-09T04:56:08.987Z"}}],"participationCoverage":{"total":5,"returned":5,"complete":true,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":null,"sourceUrl":"https://arxiv.org/abs/2611.07820","snapshotId":"1b660b59fa7bb6fdb305844b91e8cbe5f39ce1ba9480f7c9a439f43670031196","contentSha256":"f16419606dbd19b0822c24ac38473df33737086ab9120f140b898e77d7688899","textHash":"42f8da428d132245a9f8f7b672d0e61e530881d3b12c36ab9209b1cf6be640ce","observedAt":"2026-11-09T04:56:08.987Z","basis":"publisher-page-metadata","sourceLinks":45,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/ae542be642e84d180db0ae28936bd1514d68739472fe654d8675093d6d1c5d79/source"}}}],"audits":[{"publicationId":"5717f61a-94b5-447d-8aa0-b931066b6163","verdict":"uncertain","findings":[{"dimension":"attribution","status":"pass","explanation":"The five named authors match the arXiv listing. Findings, the proposal and the results are attributed to the authors rather than presented as established fact.","excerpt":"Authors:Dohoon Kim, Junho Park, Hyunwoong Bae, Seyoon Kim, Taesup Moon"},{"dimension":"summary","status":"uncertain","explanation":"The summary is faithful but hard for a non-specialist to follow. It omits the source's motivating context: LLMs are increasingly asked about a user's personal experiences. Without it, the reader is not told why memory retrieval matters. Central terms are unexplained: embedding, frozen dense embedding model, LoRA, per-query gate, R@10 and pp. Smallest repair: add the motivating sentence and add or confirm term-library tooltips for these terms. Separately, the summary repeats the source's '4-benchmark mean' while naming only three benchmarks, with no note of the mismatch. That may confuse readers and should be flagged as unclear in the source.","excerpt":"As LLMs are increasingly asked questions about a user's personal experiences, retrieval must distinguish memories by time, place, person, and event."},{"dimension":"date","status":"pass","explanation":"The originalPublished date of 2026-11-05 matches the v1 submission date. It is not the later feed date.","excerpt":"[Submitted on 5 Nov 2026]"},{"dimension":"support","status":"pass","explanation":"The claims are verbatim-supported. The significance line is labeled 'as presented' and is backed by the source's statement about training only a small fraction of parameters. The limitations and the question correctly keep benchmark gains separate from real-world reliability. No invented benefits, costs or stakeholders appear.","excerpt":"while training only a small fraction of the base parameters on a small axis-counterfactual training set"},{"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":"2026-11-09T03:53:10.211Z","model":"anthropic/claude-opus-5.5"}],"correctionUrl":"https://escapemeridian.com/corrections/"}