{"id":"02a0f6722f7d9d0156e34b5a9de012955eb7dc52a381669bf287c1616d9013da","title":"arXiv cs.AI: Query-based Submodular Data Selection for Efficient Instruction Tuning","argument":"The paper’s authors propose QSGS, a method for choosing instruction-tuning examples that aims to avoid selecting many near-duplicates. It scores examples for relevance to a target task while penalizing overlap among the selected examples. The authors also add gradient-guided synthesis, which creates synthetic examples to address target directions that the existing data does not cover. They report that, across MMLU, BBH, and TydiQA, QSGS at a 0.5% data budget matches or exceeds full-data performance on Mistral-7B and outperforms the strongest baseline across all benchmarks. The authors say diversity-aware selection brings greater benefits as the data budget shrinks. If those findings hold beyond the reported experiments, the method could make targeted fine-tuning more effective when training data is scarce; the abstract does not establish that result for other models or tasks. How well does QSGS perform on models and tasks outside the reported experiments?","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 details needed to assess the experimental setup or the reported comparisons. The abstract’s reported results concern MMLU, BBH, and TydiQA and Llama-2, Llama-3, and Mistral; it does not establish performance on other tasks or models.","speaker":{"name":"Zhao, Zhenghao, Kawanishi, Yasutomo, Zheng, Huihuo, Papka, Michael E., Yan, Yan","roleAtSource":"Article author"},"source":{"title":"Query-based Submodular Data Selection for Efficient Instruction Tuning","publisher":"arXiv cs.AI","format":"Article","url":"https://arxiv.org/abs/2611.04059","publishedAt":"2026-11-02","dateBasis":"Date basis not provided."},"evidence":[{"text":"The authors propose QSGS, which selects examples by balancing target relevance against redundancy within the selected set.","excerpt":"A modular relevance term pulls selected examples toward the target task, while a pairwise dispersion penalty suppresses intra-subset redundancy.","url":"https://arxiv.org/abs/2611.04059","relationship":"Source-stated"},{"text":"The authors add gradient-guided synthesis to generate examples for target directions that existing data does not cover.","excerpt":"identifies under-covered target directions via residual subspace analysis and generates synthetic examples to fill them.","url":"https://arxiv.org/abs/2611.04059","relationship":"Source-stated"},{"text":"The abstract reports that QSGS at a 0.5% budget matches or exceeds full-data performance on Mistral-7B across the named benchmarks, and beats the strongest baseline across all benchmarks.","excerpt":"QSGS at a 0.5% budget matches or exceeds full-data performance on Mistral-7B, outperforms the strongest baseline across all benchmarks","url":"https://arxiv.org/abs/2611.04059","relationship":"Source-stated"},{"text":"The authors report that diversity-aware selection provides increasing gains as the data budget decreases.","excerpt":"the benefit of diversity-aware selection becomes increasingly pronounced as the data budget decreases.","url":"https://arxiv.org/abs/2611.04059","relationship":"Source-stated"}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2611.04059","checkedAt":"2026-11-10T06:04:07.138Z","items":[],"note":"No suitable media was found by the supported page extraction."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"How well does QSGS perform on models and tasks outside the reported experiments?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"arXiv cs.AI: Query-based Submodular Data Selection for Efficient Instruction Tuning","summary":"The paper’s authors propose QSGS, a method for choosing instruction-tuning examples that aims to avoid selecting many near-duplicates. It scores examples for relevance to a target task while penalizing overlap among the selected examples. The authors also add gradient-guided synthesis, which creates synthetic examples to address target directions that the existing data does not cover. They report that, across MMLU, BBH, and TydiQA, QSGS at a 0.5% data budget matches or exceeds full-data performance on Mistral-7B and outperforms the strongest baseline across all benchmarks. The authors say diversity-aware selection brings greater benefits as the data budget shrinks. If those findings hold beyond the reported experiments, the method could make targeted fine-tuning more effective when training data is scarce; the abstract does not establish that result for other models or tasks. How well does QSGS perform on models and tasks outside the reported experiments?","question":"How well does QSGS perform on models and tasks outside the reported experiments?","claims":[{"text":"The authors propose QSGS, which selects examples by balancing target relevance against redundancy within the selected set.","excerpt":"A modular relevance term pulls selected examples toward the target task, while a pairwise dispersion penalty suppresses intra-subset redundancy.","support":"source-stated"},{"text":"The authors add gradient-guided synthesis to generate examples for target directions that existing data does not cover.","excerpt":"identifies under-covered target directions via residual subspace analysis and generates synthetic examples to fill them.","support":"source-stated"},{"text":"The abstract reports that QSGS at a 0.5% budget matches or exceeds full-data performance on Mistral-7B across the named benchmarks, and beats the strongest baseline across all benchmarks.","excerpt":"QSGS at a 0.5% budget matches or exceeds full-data performance on Mistral-7B, outperforms the strongest baseline across all benchmarks","support":"source-stated"},{"text":"The authors report that diversity-aware selection provides increasing gains as the data budget decreases.","excerpt":"the benefit of diversity-aware selection becomes increasingly pronounced as the data budget decreases.","support":"source-stated"}],"limitations":["The supplied evidence is the paper’s abstract, not its full methods or results; it does not provide details needed to assess the experimental setup or the reported comparisons.","The abstract’s reported results concern MMLU, BBH, and TydiQA and Llama-2, Llama-3, and Mistral; it does not establish performance on other tasks or models."],"mediaId":"17460ec05ba0a6b7b17f80b5a962fe094b713d4fc4da92f170d5b4378b39e05f","correctionReason":null,"id":"02a0f6722f7d9d0156e34b5a9de012955eb7dc52a381669bf287c1616d9013da","publicationId":"34cec497-2cc1-4f95-8768-02b3b10d7f12","version":1,"sourceUrl":"https://arxiv.org/abs/2611.04059","sourceTitle":"Query-based Submodular Data Selection for Efficient Instruction Tuning","publisher":"arXiv cs.AI","publishedAt":null,"originalPublished":{"value":"2026-11-02","precision":"day","basis":"publisher_citation_date"},"firstObservedAt":"2026-11-04T05:12:55.072Z","createdAt":"2026-11-10T05:01:31.588Z","historical":true,"entityIds":["arxiv"],"topicIds":["intelligence-and-discovery"],"media":[{"id":"17460ec05ba0a6b7b17f80b5a962fe094b713d4fc4da92f170d5b4378b39e05f","kind":"article","url":"https://arxiv.org/abs/2611.04059","sourceUrl":"https://arxiv.org/abs/2611.04059","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-11-04T06:01:36.886Z","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/2611.04059","checkedAt":"2026-11-10T06:04:07.138Z","items":[],"note":"No suitable media was found by the supported page extraction."},"editorialDecision":{"outcome":"accept","reasonCodes":["material_development"],"rubricVersion":"stream-admission-v5","mediaId":"17460ec05ba0a6b7b17f80b5a962fe094b713d4fc4da92f170d5b4378b39e05f","confidence":0.87,"provider":"typesafe","model":"jev-1.13.0","inputHash":"3b828477f9837183ae96e0b2a9e457ffa1b4d935f10566628f80dc29c68d4b80","decidedAt":"2026-11-04T06:01:37.727Z"},"provenance":{"textHash":"c56ce73df59da4604a44c8740488ee096554c96f0e83be425a9c7ab921c7f34b","observationId":"1c61a13304101cfa917d2e732d2cab45f931934a8633eac68f6b9bcc411c3b62","retrievedAt":"2026-11-04T06:01:36.775Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"4554a04d-6904-4339-815b-de93af260ad8"},"reviewer":{"model":"gpt-6-luna","runId":"c57da541-1b8b-461d-8cb3-c4da78441c27"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-10T05:01:31.588Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Zhao, Zhenghao","url":null,"kind":"unknown"},{"name":"Kawanishi, Yasutomo","url":null,"kind":"unknown"},{"name":"Zheng, Huihuo","url":null,"kind":"unknown"},{"name":"Papka, Michael E.","url":null,"kind":"unknown"},{"name":"Yan, Yan","url":null,"kind":"unknown"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Zhao, Zhenghao","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Zhao, Zhenghao","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.04059","locator":"head/meta[name=author|citation_author]"},"assertionId":"623932f9fab130c30b6691c843361981e66af387c38199e12b71793aa7e04244:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-10T06:04:07.138Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"623932f9fab130c30b6691c843361981e66af387c38199e12b71793aa7e04244","sourceUrl":"https://arxiv.org/abs/2611.04059","contentSha256":"df1b35212dbade2588f4c0860dcf39b889b6946c3a548b72161addb5265a99dc","textHash":"c56ce73df59da4604a44c8740488ee096554c96f0e83be425a9c7ab921c7f34b","observedAt":"2026-11-10T06:04:07.138Z"}},{"actor":{"name":"Kawanishi, Yasutomo","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Kawanishi, Yasutomo","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.04059","locator":"head/meta[name=author|citation_author]"},"assertionId":"623932f9fab130c30b6691c843361981e66af387c38199e12b71793aa7e04244:participation:1","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-10T06:04:07.138Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"623932f9fab130c30b6691c843361981e66af387c38199e12b71793aa7e04244","sourceUrl":"https://arxiv.org/abs/2611.04059","contentSha256":"df1b35212dbade2588f4c0860dcf39b889b6946c3a548b72161addb5265a99dc","textHash":"c56ce73df59da4604a44c8740488ee096554c96f0e83be425a9c7ab921c7f34b","observedAt":"2026-11-10T06:04:07.138Z"}},{"actor":{"name":"Zheng, Huihuo","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Zheng, Huihuo","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.04059","locator":"head/meta[name=author|citation_author]"},"assertionId":"623932f9fab130c30b6691c843361981e66af387c38199e12b71793aa7e04244:participation:2","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-10T06:04:07.138Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"623932f9fab130c30b6691c843361981e66af387c38199e12b71793aa7e04244","sourceUrl":"https://arxiv.org/abs/2611.04059","contentSha256":"df1b35212dbade2588f4c0860dcf39b889b6946c3a548b72161addb5265a99dc","textHash":"c56ce73df59da4604a44c8740488ee096554c96f0e83be425a9c7ab921c7f34b","observedAt":"2026-11-10T06:04:07.138Z"}},{"actor":{"name":"Papka, Michael E.","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Papka, Michael E.","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.04059","locator":"head/meta[name=author|citation_author]"},"assertionId":"623932f9fab130c30b6691c843361981e66af387c38199e12b71793aa7e04244:participation:3","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-10T06:04:07.138Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"623932f9fab130c30b6691c843361981e66af387c38199e12b71793aa7e04244","sourceUrl":"https://arxiv.org/abs/2611.04059","contentSha256":"df1b35212dbade2588f4c0860dcf39b889b6946c3a548b72161addb5265a99dc","textHash":"c56ce73df59da4604a44c8740488ee096554c96f0e83be425a9c7ab921c7f34b","observedAt":"2026-11-10T06:04:07.138Z"}},{"actor":{"name":"Yan, Yan","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Yan, Yan","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.04059","locator":"head/meta[name=author|citation_author]"},"assertionId":"623932f9fab130c30b6691c843361981e66af387c38199e12b71793aa7e04244:participation:4","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-10T06:04:07.138Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"623932f9fab130c30b6691c843361981e66af387c38199e12b71793aa7e04244","sourceUrl":"https://arxiv.org/abs/2611.04059","contentSha256":"df1b35212dbade2588f4c0860dcf39b889b6946c3a548b72161addb5265a99dc","textHash":"c56ce73df59da4604a44c8740488ee096554c96f0e83be425a9c7ab921c7f34b","observedAt":"2026-11-10T06:04:07.138Z"}}],"participationCoverage":{"total":5,"returned":5,"complete":true,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":null,"sourceUrl":"https://arxiv.org/abs/2611.04059","snapshotId":"623932f9fab130c30b6691c843361981e66af387c38199e12b71793aa7e04244","contentSha256":"df1b35212dbade2588f4c0860dcf39b889b6946c3a548b72161addb5265a99dc","textHash":"c56ce73df59da4604a44c8740488ee096554c96f0e83be425a9c7ab921c7f34b","observedAt":"2026-11-10T06:04:07.138Z","basis":"publisher-page-metadata","sourceLinks":43,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/02a0f6722f7d9d0156e34b5a9de012955eb7dc52a381669bf287c1616d9013da/source"}},"canonicalUrl":"https://escapemeridian.com/readings/arxiv-cs-ai-query-based-submodular-data-selection-for-efficient-02a0f672","formats":{"html":"/readings/arxiv-cs-ai-query-based-submodular-data-selection-for-efficient-02a0f672","markdown":"/readings/arxiv-cs-ai-query-based-submodular-data-selection-for-efficient-02a0f672.md","json":"/readings/arxiv-cs-ai-query-based-submodular-data-selection-for-efficient-02a0f672.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-fine-tuning","name":"Fine-tuning","explanation":"Further training of an existing model to adapt it to a particular task or use.","referenceUrl":"https://developers.google.com/machine-learning/glossary#fine-tuning"}],"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":"02a0f6722f7d9d0156e34b5a9de012955eb7dc52a381669bf287c1616d9013da","revision":"34cec497-2cc1-4f95-8768-02b3b10d7f12","revisionKind":"publication-id","title":"arXiv cs.AI: Query-based Submodular Data Selection for Efficient Instruction Tuning","url":"https://escapemeridian.com/readings/arxiv-cs-ai-query-based-submodular-data-selection-for-efficient-02a0f672"},"source":{"type":"source","id":"https://arxiv.org/abs/2611.04059","revision":"c56ce73df59da4604a44c8740488ee096554c96f0e83be425a9c7ab921c7f34b","revisionKind":"sha256-source-text","title":"Query-based Submodular Data Selection for Efficient Instruction Tuning","url":"https://arxiv.org/abs/2611.04059"},"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":"02a0f6722f7d9d0156e34b5a9de012955eb7dc52a381669bf287c1616d9013da","title":"arXiv cs.AI: Query-based Submodular Data Selection for Efficient Instruction Tuning","argument":"The paper’s authors propose QSGS, a method for choosing instruction-tuning examples that aims to avoid selecting many near-duplicates. It scores examples for relevance to a target task while penalizing overlap among the selected examples. The authors also add gradient-guided synthesis, which creates synthetic examples to address target directions that the existing data does not cover. They report that, across MMLU, BBH, and TydiQA, QSGS at a 0.5% data budget matches or exceeds full-data performance on Mistral-7B and outperforms the strongest baseline across all benchmarks. The authors say diversity-aware selection brings greater benefits as the data budget shrinks. If those findings hold beyond the reported experiments, the method could make targeted fine-tuning more effective when training data is scarce; the abstract does not establish that result for other models or tasks. How well does QSGS perform on models and tasks outside the reported experiments?","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 details needed to assess the experimental setup or the reported comparisons. The abstract’s reported results concern MMLU, BBH, and TydiQA and Llama-2, Llama-3, and Mistral; it does not establish performance on other tasks or models.","speaker":{"name":"Zhao, Zhenghao, Kawanishi, Yasutomo, Zheng, Huihuo, Papka, Michael E., Yan, Yan","roleAtSource":"Article author"},"source":{"title":"Query-based Submodular Data Selection for Efficient Instruction Tuning","publisher":"arXiv cs.AI","format":"Article","url":"https://arxiv.org/abs/2611.04059","publishedAt":"2026-11-02","dateBasis":"Date basis not provided."},"evidence":[{"text":"The authors propose QSGS, which selects examples by balancing target relevance against redundancy within the selected set.","excerpt":"A modular relevance term pulls selected examples toward the target task, while a pairwise dispersion penalty suppresses intra-subset redundancy.","url":"https://arxiv.org/abs/2611.04059","relationship":"Source-stated"},{"text":"The authors add gradient-guided synthesis to generate examples for target directions that existing data does not cover.","excerpt":"identifies under-covered target directions via residual subspace analysis and generates synthetic examples to fill them.","url":"https://arxiv.org/abs/2611.04059","relationship":"Source-stated"},{"text":"The abstract reports that QSGS at a 0.5% budget matches or exceeds full-data performance on Mistral-7B across the named benchmarks, and beats the strongest baseline across all benchmarks.","excerpt":"QSGS at a 0.5% budget matches or exceeds full-data performance on Mistral-7B, outperforms the strongest baseline across all benchmarks","url":"https://arxiv.org/abs/2611.04059","relationship":"Source-stated"},{"text":"The authors report that diversity-aware selection provides increasing gains as the data budget decreases.","excerpt":"the benefit of diversity-aware selection becomes increasingly pronounced as the data budget decreases.","url":"https://arxiv.org/abs/2611.04059","relationship":"Source-stated"}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2611.04059","checkedAt":"2026-11-10T06:04:07.138Z","items":[],"note":"No suitable media was found by the supported page extraction."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"How well does QSGS perform on models and tasks outside the reported experiments?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"arXiv cs.AI: Query-based Submodular Data Selection for Efficient Instruction Tuning","summary":"The paper’s authors propose QSGS, a method for choosing instruction-tuning examples that aims to avoid selecting many near-duplicates. It scores examples for relevance to a target task while penalizing overlap among the selected examples. The authors also add gradient-guided synthesis, which creates synthetic examples to address target directions that the existing data does not cover. They report that, across MMLU, BBH, and TydiQA, QSGS at a 0.5% data budget matches or exceeds full-data performance on Mistral-7B and outperforms the strongest baseline across all benchmarks. The authors say diversity-aware selection brings greater benefits as the data budget shrinks. If those findings hold beyond the reported experiments, the method could make targeted fine-tuning more effective when training data is scarce; the abstract does not establish that result for other models or tasks. How well does QSGS perform on models and tasks outside the reported experiments?","question":"How well does QSGS perform on models and tasks outside the reported experiments?","claims":[{"text":"The authors propose QSGS, which selects examples by balancing target relevance against redundancy within the selected set.","excerpt":"A modular relevance term pulls selected examples toward the target task, while a pairwise dispersion penalty suppresses intra-subset redundancy.","support":"source-stated"},{"text":"The authors add gradient-guided synthesis to generate examples for target directions that existing data does not cover.","excerpt":"identifies under-covered target directions via residual subspace analysis and generates synthetic examples to fill them.","support":"source-stated"},{"text":"The abstract reports that QSGS at a 0.5% budget matches or exceeds full-data performance on Mistral-7B across the named benchmarks, and beats the strongest baseline across all benchmarks.","excerpt":"QSGS at a 0.5% budget matches or exceeds full-data performance on Mistral-7B, outperforms the strongest baseline across all benchmarks","support":"source-stated"},{"text":"The authors report that diversity-aware selection provides increasing gains as the data budget decreases.","excerpt":"the benefit of diversity-aware selection becomes increasingly pronounced as the data budget decreases.","support":"source-stated"}],"limitations":["The supplied evidence is the paper’s abstract, not its full methods or results; it does not provide details needed to assess the experimental setup or the reported comparisons.","The abstract’s reported results concern MMLU, BBH, and TydiQA and Llama-2, Llama-3, and Mistral; it does not establish performance on other tasks or models."],"mediaId":"17460ec05ba0a6b7b17f80b5a962fe094b713d4fc4da92f170d5b4378b39e05f","correctionReason":null,"id":"02a0f6722f7d9d0156e34b5a9de012955eb7dc52a381669bf287c1616d9013da","publicationId":"34cec497-2cc1-4f95-8768-02b3b10d7f12","version":1,"sourceUrl":"https://arxiv.org/abs/2611.04059","sourceTitle":"Query-based Submodular Data Selection for Efficient Instruction Tuning","publisher":"arXiv cs.AI","publishedAt":null,"originalPublished":{"value":"2026-11-02","precision":"day","basis":"publisher_citation_date"},"firstObservedAt":"2026-11-04T05:12:55.072Z","createdAt":"2026-11-10T05:01:31.588Z","historical":true,"entityIds":["arxiv"],"topicIds":["intelligence-and-discovery"],"media":[{"id":"17460ec05ba0a6b7b17f80b5a962fe094b713d4fc4da92f170d5b4378b39e05f","kind":"article","url":"https://arxiv.org/abs/2611.04059","sourceUrl":"https://arxiv.org/abs/2611.04059","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-11-04T06:01:36.886Z","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/2611.04059","checkedAt":"2026-11-10T06:04:07.138Z","items":[],"note":"No suitable media was found by the supported page extraction."},"editorialDecision":{"outcome":"accept","reasonCodes":["material_development"],"rubricVersion":"stream-admission-v5","mediaId":"17460ec05ba0a6b7b17f80b5a962fe094b713d4fc4da92f170d5b4378b39e05f","confidence":0.87,"provider":"typesafe","model":"jev-1.13.0","inputHash":"3b828477f9837183ae96e0b2a9e457ffa1b4d935f10566628f80dc29c68d4b80","decidedAt":"2026-11-04T06:01:37.727Z"},"provenance":{"textHash":"c56ce73df59da4604a44c8740488ee096554c96f0e83be425a9c7ab921c7f34b","observationId":"1c61a13304101cfa917d2e732d2cab45f931934a8633eac68f6b9bcc411c3b62","retrievedAt":"2026-11-04T06:01:36.775Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"4554a04d-6904-4339-815b-de93af260ad8"},"reviewer":{"model":"gpt-6-luna","runId":"c57da541-1b8b-461d-8cb3-c4da78441c27"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-10T05:01:31.588Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Zhao, Zhenghao","url":null,"kind":"unknown"},{"name":"Kawanishi, Yasutomo","url":null,"kind":"unknown"},{"name":"Zheng, Huihuo","url":null,"kind":"unknown"},{"name":"Papka, Michael E.","url":null,"kind":"unknown"},{"name":"Yan, Yan","url":null,"kind":"unknown"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Zhao, Zhenghao","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Zhao, Zhenghao","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.04059","locator":"head/meta[name=author|citation_author]"},"assertionId":"623932f9fab130c30b6691c843361981e66af387c38199e12b71793aa7e04244:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-10T06:04:07.138Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"623932f9fab130c30b6691c843361981e66af387c38199e12b71793aa7e04244","sourceUrl":"https://arxiv.org/abs/2611.04059","contentSha256":"df1b35212dbade2588f4c0860dcf39b889b6946c3a548b72161addb5265a99dc","textHash":"c56ce73df59da4604a44c8740488ee096554c96f0e83be425a9c7ab921c7f34b","observedAt":"2026-11-10T06:04:07.138Z"}},{"actor":{"name":"Kawanishi, Yasutomo","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Kawanishi, Yasutomo","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.04059","locator":"head/meta[name=author|citation_author]"},"assertionId":"623932f9fab130c30b6691c843361981e66af387c38199e12b71793aa7e04244:participation:1","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-10T06:04:07.138Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"623932f9fab130c30b6691c843361981e66af387c38199e12b71793aa7e04244","sourceUrl":"https://arxiv.org/abs/2611.04059","contentSha256":"df1b35212dbade2588f4c0860dcf39b889b6946c3a548b72161addb5265a99dc","textHash":"c56ce73df59da4604a44c8740488ee096554c96f0e83be425a9c7ab921c7f34b","observedAt":"2026-11-10T06:04:07.138Z"}},{"actor":{"name":"Zheng, Huihuo","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Zheng, Huihuo","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.04059","locator":"head/meta[name=author|citation_author]"},"assertionId":"623932f9fab130c30b6691c843361981e66af387c38199e12b71793aa7e04244:participation:2","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-10T06:04:07.138Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"623932f9fab130c30b6691c843361981e66af387c38199e12b71793aa7e04244","sourceUrl":"https://arxiv.org/abs/2611.04059","contentSha256":"df1b35212dbade2588f4c0860dcf39b889b6946c3a548b72161addb5265a99dc","textHash":"c56ce73df59da4604a44c8740488ee096554c96f0e83be425a9c7ab921c7f34b","observedAt":"2026-11-10T06:04:07.138Z"}},{"actor":{"name":"Papka, Michael E.","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Papka, Michael E.","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.04059","locator":"head/meta[name=author|citation_author]"},"assertionId":"623932f9fab130c30b6691c843361981e66af387c38199e12b71793aa7e04244:participation:3","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-10T06:04:07.138Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"623932f9fab130c30b6691c843361981e66af387c38199e12b71793aa7e04244","sourceUrl":"https://arxiv.org/abs/2611.04059","contentSha256":"df1b35212dbade2588f4c0860dcf39b889b6946c3a548b72161addb5265a99dc","textHash":"c56ce73df59da4604a44c8740488ee096554c96f0e83be425a9c7ab921c7f34b","observedAt":"2026-11-10T06:04:07.138Z"}},{"actor":{"name":"Yan, Yan","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Yan, Yan","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.04059","locator":"head/meta[name=author|citation_author]"},"assertionId":"623932f9fab130c30b6691c843361981e66af387c38199e12b71793aa7e04244:participation:4","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-10T06:04:07.138Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"623932f9fab130c30b6691c843361981e66af387c38199e12b71793aa7e04244","sourceUrl":"https://arxiv.org/abs/2611.04059","contentSha256":"df1b35212dbade2588f4c0860dcf39b889b6946c3a548b72161addb5265a99dc","textHash":"c56ce73df59da4604a44c8740488ee096554c96f0e83be425a9c7ab921c7f34b","observedAt":"2026-11-10T06:04:07.138Z"}}],"participationCoverage":{"total":5,"returned":5,"complete":true,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":null,"sourceUrl":"https://arxiv.org/abs/2611.04059","snapshotId":"623932f9fab130c30b6691c843361981e66af387c38199e12b71793aa7e04244","contentSha256":"df1b35212dbade2588f4c0860dcf39b889b6946c3a548b72161addb5265a99dc","textHash":"c56ce73df59da4604a44c8740488ee096554c96f0e83be425a9c7ab921c7f34b","observedAt":"2026-11-10T06:04:07.138Z","basis":"publisher-page-metadata","sourceLinks":43,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/02a0f6722f7d9d0156e34b5a9de012955eb7dc52a381669bf287c1616d9013da/source"}}}],"audits":[{"publicationId":"34cec497-2cc1-4f95-8768-02b3b10d7f12","verdict":"uncertain","findings":[{"dimension":"attribution","status":"pass","explanation":"The reading attributes the method and results to the paper's authors (Zhao, Kawanishi, Zheng, Papka, Yan), uses report/say framing, and does not present the claims as established fact or EM opinion.","excerpt":"Authors:Zhenghao Zhao, Yasutomo Kawanishi, Huihuo Zheng, Michael E. Papka, Yan Yan"},{"dimension":"summary","status":"uncertain","explanation":"The reading is broadly faithful, but a non-specialist would struggle with central terms. Instruction tuning, data budget, full-data performance, target directions, and the benchmarks MMLU, BBH and TydiQA are not explained in plain language. It also omits the problem setup: a large mixed pool of training examples, a small target set describing the desired skill, and the critique that existing methods score each example independently. Smallest repair: add one sentence on that setup and the critique, and gloss the data budget as the share of the pool kept for training. Suggest term-library entries for submodular, instruction tuning and gradient-guided synthesis.","excerpt":"Given a large heterogeneous training pool and a small target set specifying the desired capability, targeted data selection aims to find a compact subset that maximizes downstream performance."},{"dimension":"date","status":"pass","explanation":"The originalPublished date of 2026-11-02 matches the arXiv v1 submission date. The feed date of Nov 4 is correctly not used as the publication date.","excerpt":"[Submitted on 2 Nov 2026]"},{"dimension":"support","status":"uncertain","explanation":"The four claims match the abstract. However, the conditional inference that the method could help 'when training data is scarce' shifts the source's condition. The abstract's setting is a large training pool under a tight selection budget, not scarce data. Smallest repair: say 'when only a small fraction of a large pool can be used for fine-tuning.' Separately, it is unclear whether the Mistral-7B full-data parity holds on each named benchmark individually.","excerpt":"the benefit of diversity-aware selection becomes increasingly pronounced as the data budget decreases"},{"dimension":"media","status":"pass","explanation":"The selected media is the verified direct arXiv abstract page for this paper, which is the most direct access to the source.","excerpt":null}],"createdAt":"2026-11-10T05:01:31.588Z","model":"anthropic/claude-opus-5.5"}],"correctionUrl":"https://escapemeridian.com/corrections/"}