{"id":"cd672fd35eb46aca5ec83baf1756cb953113dc1057a84b1c6e0b83e2ac03f374","title":"arXiv: LoRA on a Budget: Function-Aware Post-Hoc Compression for Adapter Fleets","argument":"In their paper, Shenghui Li and Thiemo voigt argue that compressing a fleet of LoRA adapters should account for how compression affects each adapter’s function, not just its spectral reconstruction. They frame the problem as allocating rank both across layers within each adapter and across adapters under a shared budget.\n\nTheir proposed Functional Rank Allocation (FRA) measures the effects of isolated module truncations, uses those measurements to search for candidate allocations, then remeasures complete candidates. It combines the resulting frontiers with a complementary spectral proposal before minimax allocation.\n\nAcross 78 adapters from three public fleets, the authors report better worst-case and tail utility than uniform spectral compression in constrained-budget regimes. They also report that isolated costs rank candidates well, but calibration shifts across proposal shapes; end-to-end verification and the spectral proposal therefore limit functional-search error.\n\nHow would FRA’s reported advantages hold up on other adapter fleets or under different evaluation conditions?","topicIds":["intelligence-and-discovery"],"domain":"science","claimType":"attributed reading","limits":"The reported comparison covers 78 adapters from three public fleets and constrained-budget regimes; the evidence does not specify how results would generalize to other fleets or evaluation conditions. The abstract does not provide detailed evaluation metrics or results for individual fleets.","speaker":{"name":"Li, Shenghui, voigt, Thiemo","roleAtSource":"Article author"},"source":{"title":"LoRA on a Budget: Function-Aware Post-Hoc Compression for Adapter Fleets","publisher":"arXiv cs.AI","format":"Article","url":"https://arxiv.org/abs/2610.36386","publishedAt":"2026-10-26","dateBasis":"Date basis not provided."},"evidence":[{"text":"The authors frame fleet compression as allocating rank both within each adapter and across the fleet under a shared budget.","excerpt":"rank must be distributed across layers within each adapter and across adapters within the fleet","url":"https://arxiv.org/abs/2610.36386","relationship":"Source-stated"},{"text":"FRA searches using measurements of isolated module truncations, then remeasures complete candidates and combines the lower-risk envelope with a spectral proposal.","excerpt":"remeasures every complete candidate, and takes the lower-risk envelope with a complementary spectral proposal","url":"https://arxiv.org/abs/2610.36386","relationship":"Source-stated"},{"text":"Across 78 adapters from three public fleets, the authors report improved worst-case and tail utility over uniform spectral compression in constrained-budget regimes.","excerpt":"Across 78 adapters from three public fleets, FRA improves worst-case and tail utility over uniform spectral compression in constrained-budget regimes.","url":"https://arxiv.org/abs/2610.36386","relationship":"Source-stated"},{"text":"The authors report that isolated costs help rank candidates, while calibration shifts across proposal shapes; verification and the spectral proposal limit functional-search error.","excerpt":"isolated costs rank candidates well, but their calibration shifts across proposal shapes; end-to-end verification and the spectral proposal therefore limit functional-search error.","url":"https://arxiv.org/abs/2610.36386","relationship":"Source-stated"}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2610.36386","checkedAt":"2026-11-14T03:20:08.313Z","items":[],"note":"No suitable media was found by the supported page extraction."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"How would FRA’s reported advantages hold up on other adapter fleets or under different evaluation conditions?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"arXiv: LoRA on a Budget: Function-Aware Post-Hoc Compression for Adapter Fleets","summary":"In their paper, Shenghui Li and Thiemo voigt argue that compressing a fleet of LoRA adapters should account for how compression affects each adapter’s function, not just its spectral reconstruction. They frame the problem as allocating rank both across layers within each adapter and across adapters under a shared budget.\n\nTheir proposed Functional Rank Allocation (FRA) measures the effects of isolated module truncations, uses those measurements to search for candidate allocations, then remeasures complete candidates. It combines the resulting frontiers with a complementary spectral proposal before minimax allocation.\n\nAcross 78 adapters from three public fleets, the authors report better worst-case and tail utility than uniform spectral compression in constrained-budget regimes. They also report that isolated costs rank candidates well, but calibration shifts across proposal shapes; end-to-end verification and the spectral proposal therefore limit functional-search error.\n\nHow would FRA’s reported advantages hold up on other adapter fleets or under different evaluation conditions?","question":"How would FRA’s reported advantages hold up on other adapter fleets or under different evaluation conditions?","claims":[{"text":"The authors frame fleet compression as allocating rank both within each adapter and across the fleet under a shared budget.","excerpt":"rank must be distributed across layers within each adapter and across adapters within the fleet","support":"source-stated"},{"text":"FRA searches using measurements of isolated module truncations, then remeasures complete candidates and combines the lower-risk envelope with a spectral proposal.","excerpt":"remeasures every complete candidate, and takes the lower-risk envelope with a complementary spectral proposal","support":"source-stated"},{"text":"Across 78 adapters from three public fleets, the authors report improved worst-case and tail utility over uniform spectral compression in constrained-budget regimes.","excerpt":"Across 78 adapters from three public fleets, FRA improves worst-case and tail utility over uniform spectral compression in constrained-budget regimes.","support":"source-stated"},{"text":"The authors report that isolated costs help rank candidates, while calibration shifts across proposal shapes; verification and the spectral proposal limit functional-search error.","excerpt":"isolated costs rank candidates well, but their calibration shifts across proposal shapes; end-to-end verification and the spectral proposal therefore limit functional-search error.","support":"source-stated"}],"limitations":["The reported comparison covers 78 adapters from three public fleets and constrained-budget regimes; the evidence does not specify how results would generalize to other fleets or evaluation conditions.","The abstract does not provide detailed evaluation metrics or results for individual fleets."],"mediaId":"1cd33358fa8ed608bec699fcbf4f590566e728cfbc5d79ffd30e077563cdab03","correctionReason":null,"id":"cd672fd35eb46aca5ec83baf1756cb953113dc1057a84b1c6e0b83e2ac03f374","publicationId":"8da01b09-54bc-4e2b-8f6a-ae0f8655bda6","version":1,"sourceUrl":"https://arxiv.org/abs/2610.36386","sourceTitle":"LoRA on a Budget: Function-Aware Post-Hoc Compression for Adapter Fleets","publisher":"arXiv cs.AI","publishedAt":null,"originalPublished":{"value":"2026-10-26","precision":"day","basis":"publisher_citation_date"},"firstObservedAt":"2026-10-28T07:50:32.417Z","createdAt":"2026-11-14T02:48:21.351Z","historical":true,"entityIds":["arxiv"],"topicIds":["intelligence-and-discovery"],"media":[{"id":"1cd33358fa8ed608bec699fcbf4f590566e728cfbc5d79ffd30e077563cdab03","kind":"article","url":"https://arxiv.org/abs/2610.36386","sourceUrl":"https://arxiv.org/abs/2610.36386","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-10-30T21:17:58.991Z","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/2610.36386","checkedAt":"2026-11-14T03:20:08.313Z","items":[],"note":"No suitable media was found by the supported page extraction."},"editorialDecision":{"outcome":"accept","reasonCodes":["material_development"],"rubricVersion":"stream-admission-v5","mediaId":"1cd33358fa8ed608bec699fcbf4f590566e728cfbc5d79ffd30e077563cdab03","confidence":0.81,"provider":"typesafe","model":"jev-1.13.0","inputHash":"1aa4ab5900a5b8ea1b0c97293fa23cff144d8fd8076a9b23c4ef24795929e04b","decidedAt":"2026-10-30T21:18:01.512Z"},"provenance":{"textHash":"c723d58a5120c5c38abec6ee9043900f2295f7c7c7a369934ec7c0b3ec600db8","observationId":"4d62dfff5e9377cd25184729f2c8bab22e63d0e4ed3ff79782fe5db732bef4b9","retrievedAt":"2026-10-30T21:17:58.627Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"66c84e1a-be6e-41b5-872a-986afc9226d4"},"reviewer":{"model":"gpt-6-luna","runId":"6b383f6f-3af5-43d3-89ca-b86146259043"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-14T02:48:21.351Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Li, Shenghui","url":null,"kind":"unknown"},{"name":"voigt, Thiemo","url":null,"kind":"unknown"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Li, Shenghui","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Li, Shenghui","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.36386","locator":"head/meta[name=author|citation_author]"},"assertionId":"9fb6e8b0c548797b7fed50c63f5dc78534255192dd7b0c2abd44b509407de986:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T21:17:58.627Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"9fb6e8b0c548797b7fed50c63f5dc78534255192dd7b0c2abd44b509407de986","sourceUrl":"https://arxiv.org/abs/2610.36386","contentSha256":"c31ce79101c239864e76346ea90f37f29268e2fe43294e88b6f02233289e380a","textHash":"c723d58a5120c5c38abec6ee9043900f2295f7c7c7a369934ec7c0b3ec600db8","observedAt":"2026-10-30T21:17:58.627Z"}},{"actor":{"name":"voigt, Thiemo","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"voigt, Thiemo","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.36386","locator":"head/meta[name=author|citation_author]"},"assertionId":"9fb6e8b0c548797b7fed50c63f5dc78534255192dd7b0c2abd44b509407de986:participation:1","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T21:17:58.627Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"9fb6e8b0c548797b7fed50c63f5dc78534255192dd7b0c2abd44b509407de986","sourceUrl":"https://arxiv.org/abs/2610.36386","contentSha256":"c31ce79101c239864e76346ea90f37f29268e2fe43294e88b6f02233289e380a","textHash":"c723d58a5120c5c38abec6ee9043900f2295f7c7c7a369934ec7c0b3ec600db8","observedAt":"2026-10-30T21:17:58.627Z"}}],"participationCoverage":{"total":2,"returned":2,"complete":true,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":null,"sourceUrl":"https://arxiv.org/abs/2610.36386","snapshotId":"9fb6e8b0c548797b7fed50c63f5dc78534255192dd7b0c2abd44b509407de986","contentSha256":"c31ce79101c239864e76346ea90f37f29268e2fe43294e88b6f02233289e380a","textHash":"c723d58a5120c5c38abec6ee9043900f2295f7c7c7a369934ec7c0b3ec600db8","observedAt":"2026-10-30T21:17:58.627Z","basis":"publisher-page-metadata","sourceLinks":40,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/cd672fd35eb46aca5ec83baf1756cb953113dc1057a84b1c6e0b83e2ac03f374/source"}},"canonicalUrl":"https://escapemeridian.com/readings/arxiv-lora-on-a-budget-function-aware-post-hoc-compression-for-cd672fd3","formats":{"html":"/readings/arxiv-lora-on-a-budget-function-aware-post-hoc-compression-for-cd672fd3","markdown":"/readings/arxiv-lora-on-a-budget-function-aware-post-hoc-compression-for-cd672fd3.md","json":"/readings/arxiv-lora-on-a-budget-function-aware-post-hoc-compression-for-cd672fd3.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":[],"conditions":[],"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":[],"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"}]},"researchQuestions":[],"withdrawn":false,"references":{"schemaVersion":1,"target":{"type":"reading","id":"cd672fd35eb46aca5ec83baf1756cb953113dc1057a84b1c6e0b83e2ac03f374","revision":"8da01b09-54bc-4e2b-8f6a-ae0f8655bda6","revisionKind":"publication-id","title":"arXiv: LoRA on a Budget: Function-Aware Post-Hoc Compression for Adapter Fleets","url":"https://escapemeridian.com/readings/arxiv-lora-on-a-budget-function-aware-post-hoc-compression-for-cd672fd3"},"source":{"type":"source","id":"https://arxiv.org/abs/2610.36386","revision":"c723d58a5120c5c38abec6ee9043900f2295f7c7c7a369934ec7c0b3ec600db8","revisionKind":"sha256-source-text","title":"LoRA on a Budget: Function-Aware Post-Hoc Compression for Adapter Fleets","url":"https://arxiv.org/abs/2610.36386"},"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":"cd672fd35eb46aca5ec83baf1756cb953113dc1057a84b1c6e0b83e2ac03f374","title":"arXiv: LoRA on a Budget: Function-Aware Post-Hoc Compression for Adapter Fleets","argument":"In their paper, Shenghui Li and Thiemo voigt argue that compressing a fleet of LoRA adapters should account for how compression affects each adapter’s function, not just its spectral reconstruction. They frame the problem as allocating rank both across layers within each adapter and across adapters under a shared budget.\n\nTheir proposed Functional Rank Allocation (FRA) measures the effects of isolated module truncations, uses those measurements to search for candidate allocations, then remeasures complete candidates. It combines the resulting frontiers with a complementary spectral proposal before minimax allocation.\n\nAcross 78 adapters from three public fleets, the authors report better worst-case and tail utility than uniform spectral compression in constrained-budget regimes. They also report that isolated costs rank candidates well, but calibration shifts across proposal shapes; end-to-end verification and the spectral proposal therefore limit functional-search error.\n\nHow would FRA’s reported advantages hold up on other adapter fleets or under different evaluation conditions?","topicIds":["intelligence-and-discovery"],"domain":"science","claimType":"attributed reading","limits":"The reported comparison covers 78 adapters from three public fleets and constrained-budget regimes; the evidence does not specify how results would generalize to other fleets or evaluation conditions. The abstract does not provide detailed evaluation metrics or results for individual fleets.","speaker":{"name":"Li, Shenghui, voigt, Thiemo","roleAtSource":"Article author"},"source":{"title":"LoRA on a Budget: Function-Aware Post-Hoc Compression for Adapter Fleets","publisher":"arXiv cs.AI","format":"Article","url":"https://arxiv.org/abs/2610.36386","publishedAt":"2026-10-26","dateBasis":"Date basis not provided."},"evidence":[{"text":"The authors frame fleet compression as allocating rank both within each adapter and across the fleet under a shared budget.","excerpt":"rank must be distributed across layers within each adapter and across adapters within the fleet","url":"https://arxiv.org/abs/2610.36386","relationship":"Source-stated"},{"text":"FRA searches using measurements of isolated module truncations, then remeasures complete candidates and combines the lower-risk envelope with a spectral proposal.","excerpt":"remeasures every complete candidate, and takes the lower-risk envelope with a complementary spectral proposal","url":"https://arxiv.org/abs/2610.36386","relationship":"Source-stated"},{"text":"Across 78 adapters from three public fleets, the authors report improved worst-case and tail utility over uniform spectral compression in constrained-budget regimes.","excerpt":"Across 78 adapters from three public fleets, FRA improves worst-case and tail utility over uniform spectral compression in constrained-budget regimes.","url":"https://arxiv.org/abs/2610.36386","relationship":"Source-stated"},{"text":"The authors report that isolated costs help rank candidates, while calibration shifts across proposal shapes; verification and the spectral proposal limit functional-search error.","excerpt":"isolated costs rank candidates well, but their calibration shifts across proposal shapes; end-to-end verification and the spectral proposal therefore limit functional-search error.","url":"https://arxiv.org/abs/2610.36386","relationship":"Source-stated"}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2610.36386","checkedAt":"2026-11-14T03:20:08.313Z","items":[],"note":"No suitable media was found by the supported page extraction."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"How would FRA’s reported advantages hold up on other adapter fleets or under different evaluation conditions?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"arXiv: LoRA on a Budget: Function-Aware Post-Hoc Compression for Adapter Fleets","summary":"In their paper, Shenghui Li and Thiemo voigt argue that compressing a fleet of LoRA adapters should account for how compression affects each adapter’s function, not just its spectral reconstruction. They frame the problem as allocating rank both across layers within each adapter and across adapters under a shared budget.\n\nTheir proposed Functional Rank Allocation (FRA) measures the effects of isolated module truncations, uses those measurements to search for candidate allocations, then remeasures complete candidates. It combines the resulting frontiers with a complementary spectral proposal before minimax allocation.\n\nAcross 78 adapters from three public fleets, the authors report better worst-case and tail utility than uniform spectral compression in constrained-budget regimes. They also report that isolated costs rank candidates well, but calibration shifts across proposal shapes; end-to-end verification and the spectral proposal therefore limit functional-search error.\n\nHow would FRA’s reported advantages hold up on other adapter fleets or under different evaluation conditions?","question":"How would FRA’s reported advantages hold up on other adapter fleets or under different evaluation conditions?","claims":[{"text":"The authors frame fleet compression as allocating rank both within each adapter and across the fleet under a shared budget.","excerpt":"rank must be distributed across layers within each adapter and across adapters within the fleet","support":"source-stated"},{"text":"FRA searches using measurements of isolated module truncations, then remeasures complete candidates and combines the lower-risk envelope with a spectral proposal.","excerpt":"remeasures every complete candidate, and takes the lower-risk envelope with a complementary spectral proposal","support":"source-stated"},{"text":"Across 78 adapters from three public fleets, the authors report improved worst-case and tail utility over uniform spectral compression in constrained-budget regimes.","excerpt":"Across 78 adapters from three public fleets, FRA improves worst-case and tail utility over uniform spectral compression in constrained-budget regimes.","support":"source-stated"},{"text":"The authors report that isolated costs help rank candidates, while calibration shifts across proposal shapes; verification and the spectral proposal limit functional-search error.","excerpt":"isolated costs rank candidates well, but their calibration shifts across proposal shapes; end-to-end verification and the spectral proposal therefore limit functional-search error.","support":"source-stated"}],"limitations":["The reported comparison covers 78 adapters from three public fleets and constrained-budget regimes; the evidence does not specify how results would generalize to other fleets or evaluation conditions.","The abstract does not provide detailed evaluation metrics or results for individual fleets."],"mediaId":"1cd33358fa8ed608bec699fcbf4f590566e728cfbc5d79ffd30e077563cdab03","correctionReason":null,"id":"cd672fd35eb46aca5ec83baf1756cb953113dc1057a84b1c6e0b83e2ac03f374","publicationId":"8da01b09-54bc-4e2b-8f6a-ae0f8655bda6","version":1,"sourceUrl":"https://arxiv.org/abs/2610.36386","sourceTitle":"LoRA on a Budget: Function-Aware Post-Hoc Compression for Adapter Fleets","publisher":"arXiv cs.AI","publishedAt":null,"originalPublished":{"value":"2026-10-26","precision":"day","basis":"publisher_citation_date"},"firstObservedAt":"2026-10-28T07:50:32.417Z","createdAt":"2026-11-14T02:48:21.351Z","historical":true,"entityIds":["arxiv"],"topicIds":["intelligence-and-discovery"],"media":[{"id":"1cd33358fa8ed608bec699fcbf4f590566e728cfbc5d79ffd30e077563cdab03","kind":"article","url":"https://arxiv.org/abs/2610.36386","sourceUrl":"https://arxiv.org/abs/2610.36386","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-10-30T21:17:58.991Z","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/2610.36386","checkedAt":"2026-11-14T03:20:08.313Z","items":[],"note":"No suitable media was found by the supported page extraction."},"editorialDecision":{"outcome":"accept","reasonCodes":["material_development"],"rubricVersion":"stream-admission-v5","mediaId":"1cd33358fa8ed608bec699fcbf4f590566e728cfbc5d79ffd30e077563cdab03","confidence":0.81,"provider":"typesafe","model":"jev-1.13.0","inputHash":"1aa4ab5900a5b8ea1b0c97293fa23cff144d8fd8076a9b23c4ef24795929e04b","decidedAt":"2026-10-30T21:18:01.512Z"},"provenance":{"textHash":"c723d58a5120c5c38abec6ee9043900f2295f7c7c7a369934ec7c0b3ec600db8","observationId":"4d62dfff5e9377cd25184729f2c8bab22e63d0e4ed3ff79782fe5db732bef4b9","retrievedAt":"2026-10-30T21:17:58.627Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"66c84e1a-be6e-41b5-872a-986afc9226d4"},"reviewer":{"model":"gpt-6-luna","runId":"6b383f6f-3af5-43d3-89ca-b86146259043"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-14T02:48:21.351Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Li, Shenghui","url":null,"kind":"unknown"},{"name":"voigt, Thiemo","url":null,"kind":"unknown"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Li, Shenghui","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Li, Shenghui","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.36386","locator":"head/meta[name=author|citation_author]"},"assertionId":"9fb6e8b0c548797b7fed50c63f5dc78534255192dd7b0c2abd44b509407de986:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T21:17:58.627Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"9fb6e8b0c548797b7fed50c63f5dc78534255192dd7b0c2abd44b509407de986","sourceUrl":"https://arxiv.org/abs/2610.36386","contentSha256":"c31ce79101c239864e76346ea90f37f29268e2fe43294e88b6f02233289e380a","textHash":"c723d58a5120c5c38abec6ee9043900f2295f7c7c7a369934ec7c0b3ec600db8","observedAt":"2026-10-30T21:17:58.627Z"}},{"actor":{"name":"voigt, Thiemo","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"voigt, Thiemo","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.36386","locator":"head/meta[name=author|citation_author]"},"assertionId":"9fb6e8b0c548797b7fed50c63f5dc78534255192dd7b0c2abd44b509407de986:participation:1","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T21:17:58.627Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"9fb6e8b0c548797b7fed50c63f5dc78534255192dd7b0c2abd44b509407de986","sourceUrl":"https://arxiv.org/abs/2610.36386","contentSha256":"c31ce79101c239864e76346ea90f37f29268e2fe43294e88b6f02233289e380a","textHash":"c723d58a5120c5c38abec6ee9043900f2295f7c7c7a369934ec7c0b3ec600db8","observedAt":"2026-10-30T21:17:58.627Z"}}],"participationCoverage":{"total":2,"returned":2,"complete":true,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":null,"sourceUrl":"https://arxiv.org/abs/2610.36386","snapshotId":"9fb6e8b0c548797b7fed50c63f5dc78534255192dd7b0c2abd44b509407de986","contentSha256":"c31ce79101c239864e76346ea90f37f29268e2fe43294e88b6f02233289e380a","textHash":"c723d58a5120c5c38abec6ee9043900f2295f7c7c7a369934ec7c0b3ec600db8","observedAt":"2026-10-30T21:17:58.627Z","basis":"publisher-page-metadata","sourceLinks":40,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/cd672fd35eb46aca5ec83baf1756cb953113dc1057a84b1c6e0b83e2ac03f374/source"}}}],"audits":[{"publicationId":"8da01b09-54bc-4e2b-8f6a-ae0f8655bda6","verdict":"uncertain","findings":[{"dimension":"attribution","status":"pass","explanation":"The reading attributes the paper and its claims to the listed authors, Shenghui Li and Thiemo voigt, with the source's capitalization preserved. Results are framed as what the authors report, not as established fact.","excerpt":"Authors:Shenghui Li, Thiemo voigt"},{"dimension":"summary","status":"uncertain","explanation":"The draft is faithful but hard for a non-specialist to follow. It never says what LoRA adapters are or why fleets of them exist, though the source says they specialize foundation models into task- and user-specific variants. It also omits that the problem arises from serving many adapters under a shared resource budget. Rank, spectral reconstruction, minimax allocation, frontiers and tail utility are left unexplained. Smallest repair: add the source's one-line purpose of LoRA and the shared serving-budget motivation, and add tooltips or plain glosses for rank, spectral and worst-case/tail utility. The title merely repeats the paper title.","excerpt":"Low-Rank Adaptation (LoRA) has become a key mechanism for efficiently specializing foundation models into many task- and user-specific variants."},{"dimension":"date","status":"pass","explanation":"originalPublished 2026-10-26 matches the v1 submission date. The later feed date of 2026-10-29 is correctly not used.","excerpt":"[Submitted on 26 Oct 2026]"},{"dimension":"support","status":"pass","explanation":"Each claim and summary sentence is grounded in the abstract. Results are scoped to 78 adapters, three public fleets and constrained-budget regimes. The limitations and the generalization question follow from that stated scope without implying unsupported benefits. Compressing the step of taking the lower-risk envelope with the spectral proposal does not change its meaning.","excerpt":"Across 78 adapters from three public fleets, FRA improves worst-case and tail utility over uniform spectral compression in constrained-budget regimes."},{"dimension":"media","status":"pass","explanation":"The selected media is the verified direct link to the arXiv abstract page, which is the original source. No other suitable media was found.","excerpt":null}],"createdAt":"2026-11-14T02:48:21.351Z","model":"anthropic/claude-opus-5.5"}],"correctionUrl":"https://escapemeridian.com/corrections/"}