{"id":"cf0c4343d93b4d1f6f0b268577e99d35ae6e5408ea967f31a4c7fd9b25ce0c53","title":"LambdaMix trains language models to shift among competing objectives","argument":"Authors Constantinos Karouzos and Nikolaos Aletras introduce LambdaMix, a method for training language models to shift among competing objectives as preferences change. They argue that standard post-training methods typically yield one policy for a fixed objective weighting. LambdaMix instead jointly trains one low-rank adapter per objective over a frozen base model; a preference vector selects the adapter mixture and weights separately normalized rewards. The authors say a single training run supports changing preferences at inference without further optimization. They experiment with four models across helpfulness and harmlessness, math problem solving while following instructions, and, for the two largest models, rewriting comments to preserve wording while reducing clues about personal attributes. On the first two tasks, they report higher hypervolume than all evaluated conditioned and policy merging baselines across all four models. They also report fewer training completions and fewer stored adapter parameters than a set of GRPO specialist policies. With architecture and initialization fixed, varying adapter mixtures and objective weights together gives wider score ranges and higher coverage than varying either alone. These are reported experimental comparisons; the abstract does not give results for the rewriting task. The paper is a preprint under review.","topicIds":["intelligence-and-discovery"],"domain":"science","claimType":"attributed reading","limits":"The paper is identified as a preprint under review. The abstract reports comparative results for the first two task groups but does not state the results for rewriting comments. The supplied abstract does not provide detailed scores or experimental procedures. This reading is limited to the abstract.","speaker":{"name":"Karouzos, Constantinos, Aletras, Nikolaos","roleAtSource":"Article author"},"source":{"title":"Steerable Multi-Objective Group-Relative Preference Optimization with Jointly Trained Low-Rank Adapters","publisher":"arXiv cs.AI","format":"Article","url":"https://arxiv.org/abs/2611.09634","publishedAt":"2026-11-06","dateBasis":"Date basis not provided."},"evidence":[{"text":"The authors introduce LambdaMix, which jointly trains one low-rank adapter per objective over a frozen base model and uses a preference vector to select adapter mixtures and weight separately normalized rewards.","excerpt":"One preference vector selects the adapter mixture that generates responses and weights the separately normalized rewards used to update the adapters.","url":"https://arxiv.org/abs/2611.09634","relationship":"Source-stated"},{"text":"The authors say one training run supports changing preferences at inference without further optimization.","excerpt":"A single training run supports changing preferences at inference without further optimization.","url":"https://arxiv.org/abs/2611.09634","relationship":"Source-stated"},{"text":"On helpfulness and harmlessness and on math problems while following instructions, LambdaMix achieves higher hypervolume than evaluated conditioned and policy merging baselines across four models.","excerpt":"On the first two tasks, LambdaMix achieves higher hypervolume than all evaluated conditioned and policy merging baselines across all four models.","url":"https://arxiv.org/abs/2611.09634","relationship":"Source-stated"},{"text":"The authors report that LambdaMix uses fewer training completions and stores fewer adapter parameters than a set of GRPO specialist policies.","excerpt":"LambdaMix uses fewer training completions and stores fewer adapter parameters than a set of GRPO specialist policies.","url":"https://arxiv.org/abs/2611.09634","relationship":"Source-stated"},{"text":"With architecture and initialization fixed, varying adapter mixtures and objective weights together gives wider score ranges and higher coverage than varying either alone.","excerpt":"With architecture and initialization fixed, varying adapter mixtures and objective weights together gives wider score ranges and higher coverage than varying either alone.","url":"https://arxiv.org/abs/2611.09634","relationship":"Source-stated"}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2611.09634","checkedAt":"2026-11-10T03:48:07.713Z","items":[],"note":"No suitable media was found by the supported page extraction."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"What results, if any, do the authors report for rewriting comments to preserve wording while reducing clues about personal attributes?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"LambdaMix trains language models to shift among competing objectives","summary":"Authors Constantinos Karouzos and Nikolaos Aletras introduce LambdaMix, a method for training language models to shift among competing objectives as preferences change. They argue that standard post-training methods typically yield one policy for a fixed objective weighting. LambdaMix instead jointly trains one low-rank adapter per objective over a frozen base model; a preference vector selects the adapter mixture and weights separately normalized rewards. The authors say a single training run supports changing preferences at inference without further optimization. They experiment with four models across helpfulness and harmlessness, math problem solving while following instructions, and, for the two largest models, rewriting comments to preserve wording while reducing clues about personal attributes. On the first two tasks, they report higher hypervolume than all evaluated conditioned and policy merging baselines across all four models. They also report fewer training completions and fewer stored adapter parameters than a set of GRPO specialist policies. With architecture and initialization fixed, varying adapter mixtures and objective weights together gives wider score ranges and higher coverage than varying either alone. These are reported experimental comparisons; the abstract does not give results for the rewriting task. The paper is a preprint under review.","question":"What results, if any, do the authors report for rewriting comments to preserve wording while reducing clues about personal attributes?","claims":[{"text":"The authors introduce LambdaMix, which jointly trains one low-rank adapter per objective over a frozen base model and uses a preference vector to select adapter mixtures and weight separately normalized rewards.","excerpt":"One preference vector selects the adapter mixture that generates responses and weights the separately normalized rewards used to update the adapters.","support":"source-stated"},{"text":"The authors say one training run supports changing preferences at inference without further optimization.","excerpt":"A single training run supports changing preferences at inference without further optimization.","support":"source-stated"},{"text":"On helpfulness and harmlessness and on math problems while following instructions, LambdaMix achieves higher hypervolume than evaluated conditioned and policy merging baselines across four models.","excerpt":"On the first two tasks, LambdaMix achieves higher hypervolume than all evaluated conditioned and policy merging baselines across all four models.","support":"source-stated"},{"text":"The authors report that LambdaMix uses fewer training completions and stores fewer adapter parameters than a set of GRPO specialist policies.","excerpt":"LambdaMix uses fewer training completions and stores fewer adapter parameters than a set of GRPO specialist policies.","support":"source-stated"},{"text":"With architecture and initialization fixed, varying adapter mixtures and objective weights together gives wider score ranges and higher coverage than varying either alone.","excerpt":"With architecture and initialization fixed, varying adapter mixtures and objective weights together gives wider score ranges and higher coverage than varying either alone.","support":"source-stated"}],"limitations":["The paper is identified as a preprint under review.","The abstract reports comparative results for the first two task groups but does not state the results for rewriting comments.","The supplied abstract does not provide detailed scores or experimental procedures. This reading is limited to the abstract."],"mediaId":"67b113e2b2f49c1106411d06cdebe09e1e13d31d5826437514bfdf51887f2382","correctionReason":null,"id":"cf0c4343d93b4d1f6f0b268577e99d35ae6e5408ea967f31a4c7fd9b25ce0c53","publicationId":"e8a1c8c8-a813-4677-851a-1006f4252580","version":1,"sourceUrl":"https://arxiv.org/abs/2611.09634","sourceTitle":"Steerable Multi-Objective Group-Relative Preference Optimization with Jointly Trained Low-Rank Adapters","publisher":"arXiv cs.AI","publishedAt":null,"originalPublished":{"value":"2026-11-06","precision":"day","basis":"publisher_citation_date"},"firstObservedAt":"2026-11-09T05:04:20.020Z","createdAt":"2026-11-10T03:16:36.404Z","historical":true,"entityIds":["arxiv"],"topicIds":["intelligence-and-discovery"],"media":[{"id":"67b113e2b2f49c1106411d06cdebe09e1e13d31d5826437514bfdf51887f2382","kind":"article","url":"https://arxiv.org/abs/2611.09634","sourceUrl":"https://arxiv.org/abs/2611.09634","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-11-10T00:02:32.789Z","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/2611.09634","checkedAt":"2026-11-10T03:48:07.713Z","items":[],"note":"No suitable media was found by the supported page extraction."},"editorialDecision":{"outcome":"accept","reasonCodes":["material_development"],"rubricVersion":"stream-admission-v5","mediaId":"67b113e2b2f49c1106411d06cdebe09e1e13d31d5826437514bfdf51887f2382","confidence":0.85,"provider":"typesafe","model":"jev-1.13.0","inputHash":"1abbb2a37e084ca4fca705211f52139585729e2b91308f8c4e9255b93ae3c91c","decidedAt":"2026-11-10T00:02:34.917Z"},"provenance":{"textHash":"9aa25db1c3227f448fe389e548a6fa4dd90775441f6ed8f909e431fc19d54913","observationId":"8f9b2c8ef684d0ef2df1f949dd815e37e771aadafa9e41224a1b8890f55e185e","retrievedAt":"2026-11-10T00:02:32.457Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"1ec09def-2012-433c-8a35-eeb928bc15ed"},"reviewer":{"model":"gpt-6-luna","runId":"64d69ca8-b07a-4daa-851d-ce70b45bfe27"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-10T03:16:36.404Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Karouzos, Constantinos","url":null,"kind":"unknown"},{"name":"Aletras, Nikolaos","url":null,"kind":"unknown"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Karouzos, Constantinos","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Karouzos, Constantinos","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.09634","locator":"head/meta[name=author|citation_author]"},"assertionId":"d947b6f6309e4bf4e5cbf7eadae2b80e9fc735fb5f145cd2e3d6b9d20b468da8:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-10T03:48:07.713Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"d947b6f6309e4bf4e5cbf7eadae2b80e9fc735fb5f145cd2e3d6b9d20b468da8","sourceUrl":"https://arxiv.org/abs/2611.09634","contentSha256":"32b238acc70461909354db0c58a6e280f461d082c83d6433de3019e3b97fe8dd","textHash":"9aa25db1c3227f448fe389e548a6fa4dd90775441f6ed8f909e431fc19d54913","observedAt":"2026-11-10T03:48:07.713Z"}},{"actor":{"name":"Aletras, Nikolaos","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Aletras, Nikolaos","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.09634","locator":"head/meta[name=author|citation_author]"},"assertionId":"d947b6f6309e4bf4e5cbf7eadae2b80e9fc735fb5f145cd2e3d6b9d20b468da8:participation:1","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-10T03:48:07.713Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"d947b6f6309e4bf4e5cbf7eadae2b80e9fc735fb5f145cd2e3d6b9d20b468da8","sourceUrl":"https://arxiv.org/abs/2611.09634","contentSha256":"32b238acc70461909354db0c58a6e280f461d082c83d6433de3019e3b97fe8dd","textHash":"9aa25db1c3227f448fe389e548a6fa4dd90775441f6ed8f909e431fc19d54913","observedAt":"2026-11-10T03:48:07.713Z"}}],"participationCoverage":{"total":2,"returned":2,"complete":true,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":null,"sourceUrl":"https://arxiv.org/abs/2611.09634","snapshotId":"d947b6f6309e4bf4e5cbf7eadae2b80e9fc735fb5f145cd2e3d6b9d20b468da8","contentSha256":"32b238acc70461909354db0c58a6e280f461d082c83d6433de3019e3b97fe8dd","textHash":"9aa25db1c3227f448fe389e548a6fa4dd90775441f6ed8f909e431fc19d54913","observedAt":"2026-11-10T03:48:07.713Z","basis":"publisher-page-metadata","sourceLinks":42,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/cf0c4343d93b4d1f6f0b268577e99d35ae6e5408ea967f31a4c7fd9b25ce0c53/source"}},"canonicalUrl":"https://escapemeridian.com/readings/lambdamix-trains-language-models-to-shift-among-competing-cf0c4343","formats":{"html":"/readings/lambdamix-trains-language-models-to-shift-among-competing-cf0c4343","markdown":"/readings/lambdamix-trains-language-models-to-shift-among-competing-cf0c4343.md","json":"/readings/lambdamix-trains-language-models-to-shift-among-competing-cf0c4343.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":[],"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":"cf0c4343d93b4d1f6f0b268577e99d35ae6e5408ea967f31a4c7fd9b25ce0c53","revision":"e8a1c8c8-a813-4677-851a-1006f4252580","revisionKind":"publication-id","title":"LambdaMix trains language models to shift among competing objectives","url":"https://escapemeridian.com/readings/lambdamix-trains-language-models-to-shift-among-competing-cf0c4343"},"source":{"type":"source","id":"https://arxiv.org/abs/2611.09634","revision":"9aa25db1c3227f448fe389e548a6fa4dd90775441f6ed8f909e431fc19d54913","revisionKind":"sha256-source-text","title":"Steerable Multi-Objective Group-Relative Preference Optimization with Jointly Trained Low-Rank Adapters","url":"https://arxiv.org/abs/2611.09634"},"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":"cf0c4343d93b4d1f6f0b268577e99d35ae6e5408ea967f31a4c7fd9b25ce0c53","title":"LambdaMix trains language models to shift among competing objectives","argument":"Authors Constantinos Karouzos and Nikolaos Aletras introduce LambdaMix, a method for training language models to shift among competing objectives as preferences change. They argue that standard post-training methods typically yield one policy for a fixed objective weighting. LambdaMix instead jointly trains one low-rank adapter per objective over a frozen base model; a preference vector selects the adapter mixture and weights separately normalized rewards. The authors say a single training run supports changing preferences at inference without further optimization. They experiment with four models across helpfulness and harmlessness, math problem solving while following instructions, and, for the two largest models, rewriting comments to preserve wording while reducing clues about personal attributes. On the first two tasks, they report higher hypervolume than all evaluated conditioned and policy merging baselines across all four models. They also report fewer training completions and fewer stored adapter parameters than a set of GRPO specialist policies. With architecture and initialization fixed, varying adapter mixtures and objective weights together gives wider score ranges and higher coverage than varying either alone. These are reported experimental comparisons; the abstract does not give results for the rewriting task. The paper is a preprint under review.","topicIds":["intelligence-and-discovery"],"domain":"science","claimType":"attributed reading","limits":"The paper is identified as a preprint under review. The abstract reports comparative results for the first two task groups but does not state the results for rewriting comments. The supplied abstract does not provide detailed scores or experimental procedures. This reading is limited to the abstract.","speaker":{"name":"Karouzos, Constantinos, Aletras, Nikolaos","roleAtSource":"Article author"},"source":{"title":"Steerable Multi-Objective Group-Relative Preference Optimization with Jointly Trained Low-Rank Adapters","publisher":"arXiv cs.AI","format":"Article","url":"https://arxiv.org/abs/2611.09634","publishedAt":"2026-11-06","dateBasis":"Date basis not provided."},"evidence":[{"text":"The authors introduce LambdaMix, which jointly trains one low-rank adapter per objective over a frozen base model and uses a preference vector to select adapter mixtures and weight separately normalized rewards.","excerpt":"One preference vector selects the adapter mixture that generates responses and weights the separately normalized rewards used to update the adapters.","url":"https://arxiv.org/abs/2611.09634","relationship":"Source-stated"},{"text":"The authors say one training run supports changing preferences at inference without further optimization.","excerpt":"A single training run supports changing preferences at inference without further optimization.","url":"https://arxiv.org/abs/2611.09634","relationship":"Source-stated"},{"text":"On helpfulness and harmlessness and on math problems while following instructions, LambdaMix achieves higher hypervolume than evaluated conditioned and policy merging baselines across four models.","excerpt":"On the first two tasks, LambdaMix achieves higher hypervolume than all evaluated conditioned and policy merging baselines across all four models.","url":"https://arxiv.org/abs/2611.09634","relationship":"Source-stated"},{"text":"The authors report that LambdaMix uses fewer training completions and stores fewer adapter parameters than a set of GRPO specialist policies.","excerpt":"LambdaMix uses fewer training completions and stores fewer adapter parameters than a set of GRPO specialist policies.","url":"https://arxiv.org/abs/2611.09634","relationship":"Source-stated"},{"text":"With architecture and initialization fixed, varying adapter mixtures and objective weights together gives wider score ranges and higher coverage than varying either alone.","excerpt":"With architecture and initialization fixed, varying adapter mixtures and objective weights together gives wider score ranges and higher coverage than varying either alone.","url":"https://arxiv.org/abs/2611.09634","relationship":"Source-stated"}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2611.09634","checkedAt":"2026-11-10T03:48:07.713Z","items":[],"note":"No suitable media was found by the supported page extraction."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"What results, if any, do the authors report for rewriting comments to preserve wording while reducing clues about personal attributes?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"LambdaMix trains language models to shift among competing objectives","summary":"Authors Constantinos Karouzos and Nikolaos Aletras introduce LambdaMix, a method for training language models to shift among competing objectives as preferences change. They argue that standard post-training methods typically yield one policy for a fixed objective weighting. LambdaMix instead jointly trains one low-rank adapter per objective over a frozen base model; a preference vector selects the adapter mixture and weights separately normalized rewards. The authors say a single training run supports changing preferences at inference without further optimization. They experiment with four models across helpfulness and harmlessness, math problem solving while following instructions, and, for the two largest models, rewriting comments to preserve wording while reducing clues about personal attributes. On the first two tasks, they report higher hypervolume than all evaluated conditioned and policy merging baselines across all four models. They also report fewer training completions and fewer stored adapter parameters than a set of GRPO specialist policies. With architecture and initialization fixed, varying adapter mixtures and objective weights together gives wider score ranges and higher coverage than varying either alone. These are reported experimental comparisons; the abstract does not give results for the rewriting task. The paper is a preprint under review.","question":"What results, if any, do the authors report for rewriting comments to preserve wording while reducing clues about personal attributes?","claims":[{"text":"The authors introduce LambdaMix, which jointly trains one low-rank adapter per objective over a frozen base model and uses a preference vector to select adapter mixtures and weight separately normalized rewards.","excerpt":"One preference vector selects the adapter mixture that generates responses and weights the separately normalized rewards used to update the adapters.","support":"source-stated"},{"text":"The authors say one training run supports changing preferences at inference without further optimization.","excerpt":"A single training run supports changing preferences at inference without further optimization.","support":"source-stated"},{"text":"On helpfulness and harmlessness and on math problems while following instructions, LambdaMix achieves higher hypervolume than evaluated conditioned and policy merging baselines across four models.","excerpt":"On the first two tasks, LambdaMix achieves higher hypervolume than all evaluated conditioned and policy merging baselines across all four models.","support":"source-stated"},{"text":"The authors report that LambdaMix uses fewer training completions and stores fewer adapter parameters than a set of GRPO specialist policies.","excerpt":"LambdaMix uses fewer training completions and stores fewer adapter parameters than a set of GRPO specialist policies.","support":"source-stated"},{"text":"With architecture and initialization fixed, varying adapter mixtures and objective weights together gives wider score ranges and higher coverage than varying either alone.","excerpt":"With architecture and initialization fixed, varying adapter mixtures and objective weights together gives wider score ranges and higher coverage than varying either alone.","support":"source-stated"}],"limitations":["The paper is identified as a preprint under review.","The abstract reports comparative results for the first two task groups but does not state the results for rewriting comments.","The supplied abstract does not provide detailed scores or experimental procedures. This reading is limited to the abstract."],"mediaId":"67b113e2b2f49c1106411d06cdebe09e1e13d31d5826437514bfdf51887f2382","correctionReason":null,"id":"cf0c4343d93b4d1f6f0b268577e99d35ae6e5408ea967f31a4c7fd9b25ce0c53","publicationId":"e8a1c8c8-a813-4677-851a-1006f4252580","version":1,"sourceUrl":"https://arxiv.org/abs/2611.09634","sourceTitle":"Steerable Multi-Objective Group-Relative Preference Optimization with Jointly Trained Low-Rank Adapters","publisher":"arXiv cs.AI","publishedAt":null,"originalPublished":{"value":"2026-11-06","precision":"day","basis":"publisher_citation_date"},"firstObservedAt":"2026-11-09T05:04:20.020Z","createdAt":"2026-11-10T03:16:36.404Z","historical":true,"entityIds":["arxiv"],"topicIds":["intelligence-and-discovery"],"media":[{"id":"67b113e2b2f49c1106411d06cdebe09e1e13d31d5826437514bfdf51887f2382","kind":"article","url":"https://arxiv.org/abs/2611.09634","sourceUrl":"https://arxiv.org/abs/2611.09634","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-11-10T00:02:32.789Z","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/2611.09634","checkedAt":"2026-11-10T03:48:07.713Z","items":[],"note":"No suitable media was found by the supported page extraction."},"editorialDecision":{"outcome":"accept","reasonCodes":["material_development"],"rubricVersion":"stream-admission-v5","mediaId":"67b113e2b2f49c1106411d06cdebe09e1e13d31d5826437514bfdf51887f2382","confidence":0.85,"provider":"typesafe","model":"jev-1.13.0","inputHash":"1abbb2a37e084ca4fca705211f52139585729e2b91308f8c4e9255b93ae3c91c","decidedAt":"2026-11-10T00:02:34.917Z"},"provenance":{"textHash":"9aa25db1c3227f448fe389e548a6fa4dd90775441f6ed8f909e431fc19d54913","observationId":"8f9b2c8ef684d0ef2df1f949dd815e37e771aadafa9e41224a1b8890f55e185e","retrievedAt":"2026-11-10T00:02:32.457Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"1ec09def-2012-433c-8a35-eeb928bc15ed"},"reviewer":{"model":"gpt-6-luna","runId":"64d69ca8-b07a-4daa-851d-ce70b45bfe27"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-10T03:16:36.404Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Karouzos, Constantinos","url":null,"kind":"unknown"},{"name":"Aletras, Nikolaos","url":null,"kind":"unknown"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Karouzos, Constantinos","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Karouzos, Constantinos","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.09634","locator":"head/meta[name=author|citation_author]"},"assertionId":"d947b6f6309e4bf4e5cbf7eadae2b80e9fc735fb5f145cd2e3d6b9d20b468da8:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-10T03:48:07.713Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"d947b6f6309e4bf4e5cbf7eadae2b80e9fc735fb5f145cd2e3d6b9d20b468da8","sourceUrl":"https://arxiv.org/abs/2611.09634","contentSha256":"32b238acc70461909354db0c58a6e280f461d082c83d6433de3019e3b97fe8dd","textHash":"9aa25db1c3227f448fe389e548a6fa4dd90775441f6ed8f909e431fc19d54913","observedAt":"2026-11-10T03:48:07.713Z"}},{"actor":{"name":"Aletras, Nikolaos","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Aletras, Nikolaos","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.09634","locator":"head/meta[name=author|citation_author]"},"assertionId":"d947b6f6309e4bf4e5cbf7eadae2b80e9fc735fb5f145cd2e3d6b9d20b468da8:participation:1","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-10T03:48:07.713Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"d947b6f6309e4bf4e5cbf7eadae2b80e9fc735fb5f145cd2e3d6b9d20b468da8","sourceUrl":"https://arxiv.org/abs/2611.09634","contentSha256":"32b238acc70461909354db0c58a6e280f461d082c83d6433de3019e3b97fe8dd","textHash":"9aa25db1c3227f448fe389e548a6fa4dd90775441f6ed8f909e431fc19d54913","observedAt":"2026-11-10T03:48:07.713Z"}}],"participationCoverage":{"total":2,"returned":2,"complete":true,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":null,"sourceUrl":"https://arxiv.org/abs/2611.09634","snapshotId":"d947b6f6309e4bf4e5cbf7eadae2b80e9fc735fb5f145cd2e3d6b9d20b468da8","contentSha256":"32b238acc70461909354db0c58a6e280f461d082c83d6433de3019e3b97fe8dd","textHash":"9aa25db1c3227f448fe389e548a6fa4dd90775441f6ed8f909e431fc19d54913","observedAt":"2026-11-10T03:48:07.713Z","basis":"publisher-page-metadata","sourceLinks":42,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/cf0c4343d93b4d1f6f0b268577e99d35ae6e5408ea967f31a4c7fd9b25ce0c53/source"}}}],"audits":[{"publicationId":"e8a1c8c8-a813-4677-851a-1006f4252580","verdict":"uncertain","findings":[{"dimension":"attribution","status":"uncertain","explanation":"The authors, Constantinos Karouzos and Nikolaos Aletras, are correctly named, and most results are attributed. However, the summary sentence beginning 'With architecture and initialization fixed...' and claim 5 state the ablation finding as plain fact. It should stay framed as the authors' report. Smallest repair: prefix it with 'They also report that'.","excerpt":"With architecture and initialization fixed, varying adapter mixtures and objective weights together gives wider score ranges and higher coverage than varying either alone."},{"dimension":"summary","status":"uncertain","explanation":"The reading is faithful, but a non-specialist cannot follow it. Central terms are used without any plain-language explanation or term-library tooltip: post-training, policy, low-rank adapter, frozen base model, preference vector, hypervolume, conditioned and policy merging baselines, GRPO and training completions. The reading also never explains why one steerable model, rather than several specialists, matters, for example lower training and storage cost. Repair: add tooltips for these terms and one plain sentence on that practical significance. The title and question are fine.","excerpt":"standard post-training methods typically produce a single policy for a fixed objective weighting"},{"dimension":"date","status":"pass","explanation":"originalPublished is 2026-11-06. This matches the v1 submission date on arXiv. The later feed date of Nov 9 is not used as the original date.","excerpt":"[Submitted on 6 Nov 2026]"},{"dimension":"support","status":"pass","explanation":"All claims match the abstract nearly verbatim. The limitations are accurate: the abstract gives no rewriting-task results, the paper is a preprint under review, and no detailed scores are given. The question follows from a real gap in the evidence. No benefits or costs are invented beyond what the source reports.","excerpt":"On the first two tasks, LambdaMix achieves higher hypervolume than all evaluated conditioned and policy merging baselines across all four models."},{"dimension":"media","status":"pass","explanation":"The selected media is the verified direct link to the arXiv abstract page of the source itself.","excerpt":null}],"createdAt":"2026-11-10T03:16:36.404Z","model":"anthropic/claude-opus-5.5"}],"correctionUrl":"https://escapemeridian.com/corrections/"}