{"id":"e2babda51d94af28b41b581a7d73079d9cc83c7e82c3ef7dda44dc18e661bfd0","title":"The paper argues OPD learns where reasoning paths stay open","argument":"Authors Shiping Yang, Chuanqi Tang, Hossein Hajimirsadeghi, Layla El Asri, Mohammad Amin Shabani, and Angel X. Chang argue that on-policy distillation (OPD) learns more than which candidate tokens a student considers. It also learns where the student keeps reasoning paths open, which the authors call the “shape” of uncertainty.\n\nThe authors introduce entropy shape—the sequence of next-token entropies—and Entropy Shape Correlation (ESC), a measure of student-teacher agreement on that sequence. They report that intervening on a teacher’s entropy while preserving its token preferences indicates that where to branch causally drives OPD gains, independently of how to branch.\n\nThey also frame predicting OPD outcomes as a teacher-selection task. In their evaluation protocol, pre-OPD ESC selects the best teacher for both students and outperforms existing predictors. If this measure helps identify a useful teacher before distillation, it could make teacher selection more informed; the abstract does not establish how well the result transfers beyond that protocol. What results would show whether pre-OPD ESC predicts teacher quality for students and settings outside the authors’ evaluation protocol?","topicIds":["intelligence-and-discovery"],"domain":"science","claimType":"attributed reading","limits":"The evidence is the paper’s abstract; it does not provide the evaluation protocol’s details, measurements, or the scope of students and settings tested. The abstract does not establish whether pre-OPD ESC predicts teacher quality beyond the authors’ evaluation protocol.","speaker":{"name":"Yang, Shiping, Tang, Chuanqi, Hajimirsadeghi, Hossein, Asri, Layla El, Shabani, Mohammad Amin, Chang, Angel X.","roleAtSource":"Article author"},"source":{"title":"Distilling the Shape of Uncertainty: What On-Policy Distillation Learns and How to Predict It","publisher":"arXiv cs.AI","format":"Article","url":"https://arxiv.org/abs/2610.39020","publishedAt":"2026-10-28","dateBasis":"Date basis not provided."},"evidence":[{"text":"The authors argue that OPD learns both which candidate tokens to consider and where to keep reasoning paths open.","excerpt":"OPD learns not only \\textit{how to branch}, i.e., which candidate tokens to consider, but also \\textit{where to branch}, i.e., where to keep reasoning paths open.","url":"https://arxiv.org/abs/2610.39020","relationship":"Source-stated"},{"text":"The authors introduce entropy shape and ESC to characterize and measure agreement on where reasoning paths remain open.","excerpt":"We introduce the entropy shape, the sequence of next-token entropies, to characterize the latter, and propose the Entropy Shape Correlation (ESC) to measure student-teacher agreement in entropy shape.","url":"https://arxiv.org/abs/2610.39020","relationship":"Source-stated"},{"text":"The authors report that where to branch causally drives OPD gains independently of how to branch, based on tracking ESC and intervening on teacher entropy while preserving token preferences.","excerpt":"intervening on a teacher's entropy while preserving its token preferences, we empirically show that \\textit{where to branch} causally drives OPD gains, independently of \\textit{how to branch}.","url":"https://arxiv.org/abs/2610.39020","relationship":"Source-stated"},{"text":"Under the authors’ evaluation protocol, pre-OPD ESC selects the best teacher for both students and outperforms existing predictors.","excerpt":"under which pre-OPD ESC selects the best teacher for both students and outperforms existing predictors.","url":"https://arxiv.org/abs/2610.39020","relationship":"Source-stated"}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2610.39020","checkedAt":"2026-11-11T08:56:07.537Z","items":[],"note":"No suitable media was found by the supported page extraction."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"What results would show whether pre-OPD ESC predicts teacher quality for students and settings outside the authors’ evaluation protocol?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"The paper argues OPD learns where reasoning paths stay open","summary":"Authors Shiping Yang, Chuanqi Tang, Hossein Hajimirsadeghi, Layla El Asri, Mohammad Amin Shabani, and Angel X. Chang argue that on-policy distillation (OPD) learns more than which candidate tokens a student considers. It also learns where the student keeps reasoning paths open, which the authors call the “shape” of uncertainty.\n\nThe authors introduce entropy shape—the sequence of next-token entropies—and Entropy Shape Correlation (ESC), a measure of student-teacher agreement on that sequence. They report that intervening on a teacher’s entropy while preserving its token preferences indicates that where to branch causally drives OPD gains, independently of how to branch.\n\nThey also frame predicting OPD outcomes as a teacher-selection task. In their evaluation protocol, pre-OPD ESC selects the best teacher for both students and outperforms existing predictors. If this measure helps identify a useful teacher before distillation, it could make teacher selection more informed; the abstract does not establish how well the result transfers beyond that protocol. What results would show whether pre-OPD ESC predicts teacher quality for students and settings outside the authors’ evaluation protocol?","question":"What results would show whether pre-OPD ESC predicts teacher quality for students and settings outside the authors’ evaluation protocol?","claims":[{"text":"The authors argue that OPD learns both which candidate tokens to consider and where to keep reasoning paths open.","excerpt":"OPD learns not only \\textit{how to branch}, i.e., which candidate tokens to consider, but also \\textit{where to branch}, i.e., where to keep reasoning paths open.","support":"source-stated"},{"text":"The authors introduce entropy shape and ESC to characterize and measure agreement on where reasoning paths remain open.","excerpt":"We introduce the entropy shape, the sequence of next-token entropies, to characterize the latter, and propose the Entropy Shape Correlation (ESC) to measure student-teacher agreement in entropy shape.","support":"source-stated"},{"text":"The authors report that where to branch causally drives OPD gains independently of how to branch, based on tracking ESC and intervening on teacher entropy while preserving token preferences.","excerpt":"intervening on a teacher's entropy while preserving its token preferences, we empirically show that \\textit{where to branch} causally drives OPD gains, independently of \\textit{how to branch}.","support":"source-stated"},{"text":"Under the authors’ evaluation protocol, pre-OPD ESC selects the best teacher for both students and outperforms existing predictors.","excerpt":"under which pre-OPD ESC selects the best teacher for both students and outperforms existing predictors.","support":"source-stated"}],"limitations":["The evidence is the paper’s abstract; it does not provide the evaluation protocol’s details, measurements, or the scope of students and settings tested.","The abstract does not establish whether pre-OPD ESC predicts teacher quality beyond the authors’ evaluation protocol."],"mediaId":"f4a7a7198e1aa606bea8a43ac56f9a1c999a99327ce8148192af7551aacc86d6","correctionReason":null,"id":"e2babda51d94af28b41b581a7d73079d9cc83c7e82c3ef7dda44dc18e661bfd0","publicationId":"daadfca5-7868-4fbd-8892-e560800eca6c","version":1,"sourceUrl":"https://arxiv.org/abs/2610.39020","sourceTitle":"Distilling the Shape of Uncertainty: What On-Policy Distillation Learns and How to Predict It","publisher":"arXiv cs.AI","publishedAt":null,"originalPublished":{"value":"2026-10-28","precision":"day","basis":"publisher_citation_date"},"firstObservedAt":"2026-10-30T06:42:44.412Z","createdAt":"2026-11-11T08:25:25.106Z","historical":true,"entityIds":["arxiv"],"topicIds":["intelligence-and-discovery"],"media":[{"id":"f4a7a7198e1aa606bea8a43ac56f9a1c999a99327ce8148192af7551aacc86d6","kind":"article","url":"https://arxiv.org/abs/2610.39020","sourceUrl":"https://arxiv.org/abs/2610.39020","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-10-31T11:46:50.525Z","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/2610.39020","checkedAt":"2026-11-11T08:56:07.537Z","items":[],"note":"No suitable media was found by the supported page extraction."},"editorialDecision":{"outcome":"accept","reasonCodes":["material_development"],"rubricVersion":"stream-admission-v5","mediaId":"f4a7a7198e1aa606bea8a43ac56f9a1c999a99327ce8148192af7551aacc86d6","confidence":0.8400000000000001,"provider":"typesafe","model":"jev-1.13.0","inputHash":"65a29790ce85216f1ed79c8d1ca188c384e454af5f4a9dc7b89e5ae7555944f8","decidedAt":"2026-10-31T11:46:52.237Z"},"provenance":{"textHash":"7d0962984b7e69e4c87e6ca815ffb254ae30c6566ba55c1ff1dc4758024ece5c","observationId":"4858c3e836a9c21eb1834f6ebced66f1b13a4ed4b8497895227608cc71448934","retrievedAt":"2026-10-31T11:46:50.280Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"5bd6b3ff-0559-44da-83d4-bee6193ab45f"},"reviewer":{"model":"gpt-6-luna","runId":"1a7e2cb5-e4b6-41f7-8112-9a6ceb2dbb78"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-11T08:25:25.106Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Yang, Shiping","url":null,"kind":"unknown"},{"name":"Tang, Chuanqi","url":null,"kind":"unknown"},{"name":"Hajimirsadeghi, Hossein","url":null,"kind":"unknown"},{"name":"Asri, Layla El","url":null,"kind":"unknown"},{"name":"Shabani, Mohammad Amin","url":null,"kind":"unknown"},{"name":"Chang, Angel X.","url":null,"kind":"unknown"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Yang, Shiping","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Yang, Shiping","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.39020","locator":"head/meta[name=author|citation_author]"},"assertionId":"29017aa2c9041f9708a2880ffabc45c58c2045e13de90200e74f182092c7d9ed:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-11T08:56:07.537Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"29017aa2c9041f9708a2880ffabc45c58c2045e13de90200e74f182092c7d9ed","sourceUrl":"https://arxiv.org/abs/2610.39020","contentSha256":"dc8bbb425111e3b414ffe63f69edcaf0257ca053ac60d2cbb1d8d94f2a4e6b59","textHash":"7d0962984b7e69e4c87e6ca815ffb254ae30c6566ba55c1ff1dc4758024ece5c","observedAt":"2026-11-11T08:56:07.537Z"}},{"actor":{"name":"Tang, Chuanqi","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Tang, Chuanqi","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.39020","locator":"head/meta[name=author|citation_author]"},"assertionId":"29017aa2c9041f9708a2880ffabc45c58c2045e13de90200e74f182092c7d9ed:participation:1","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-11T08:56:07.537Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"29017aa2c9041f9708a2880ffabc45c58c2045e13de90200e74f182092c7d9ed","sourceUrl":"https://arxiv.org/abs/2610.39020","contentSha256":"dc8bbb425111e3b414ffe63f69edcaf0257ca053ac60d2cbb1d8d94f2a4e6b59","textHash":"7d0962984b7e69e4c87e6ca815ffb254ae30c6566ba55c1ff1dc4758024ece5c","observedAt":"2026-11-11T08:56:07.537Z"}},{"actor":{"name":"Hajimirsadeghi, Hossein","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Hajimirsadeghi, Hossein","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.39020","locator":"head/meta[name=author|citation_author]"},"assertionId":"29017aa2c9041f9708a2880ffabc45c58c2045e13de90200e74f182092c7d9ed:participation:2","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-11T08:56:07.537Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"29017aa2c9041f9708a2880ffabc45c58c2045e13de90200e74f182092c7d9ed","sourceUrl":"https://arxiv.org/abs/2610.39020","contentSha256":"dc8bbb425111e3b414ffe63f69edcaf0257ca053ac60d2cbb1d8d94f2a4e6b59","textHash":"7d0962984b7e69e4c87e6ca815ffb254ae30c6566ba55c1ff1dc4758024ece5c","observedAt":"2026-11-11T08:56:07.537Z"}},{"actor":{"name":"Asri, Layla El","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Asri, Layla El","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.39020","locator":"head/meta[name=author|citation_author]"},"assertionId":"29017aa2c9041f9708a2880ffabc45c58c2045e13de90200e74f182092c7d9ed:participation:3","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-11T08:56:07.537Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"29017aa2c9041f9708a2880ffabc45c58c2045e13de90200e74f182092c7d9ed","sourceUrl":"https://arxiv.org/abs/2610.39020","contentSha256":"dc8bbb425111e3b414ffe63f69edcaf0257ca053ac60d2cbb1d8d94f2a4e6b59","textHash":"7d0962984b7e69e4c87e6ca815ffb254ae30c6566ba55c1ff1dc4758024ece5c","observedAt":"2026-11-11T08:56:07.537Z"}},{"actor":{"name":"Shabani, Mohammad Amin","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Shabani, Mohammad Amin","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.39020","locator":"head/meta[name=author|citation_author]"},"assertionId":"29017aa2c9041f9708a2880ffabc45c58c2045e13de90200e74f182092c7d9ed:participation:4","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-11T08:56:07.537Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"29017aa2c9041f9708a2880ffabc45c58c2045e13de90200e74f182092c7d9ed","sourceUrl":"https://arxiv.org/abs/2610.39020","contentSha256":"dc8bbb425111e3b414ffe63f69edcaf0257ca053ac60d2cbb1d8d94f2a4e6b59","textHash":"7d0962984b7e69e4c87e6ca815ffb254ae30c6566ba55c1ff1dc4758024ece5c","observedAt":"2026-11-11T08:56:07.537Z"}},{"actor":{"name":"Chang, Angel X.","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Chang, Angel X.","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.39020","locator":"head/meta[name=author|citation_author]"},"assertionId":"29017aa2c9041f9708a2880ffabc45c58c2045e13de90200e74f182092c7d9ed:participation:5","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-11T08:56:07.537Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"29017aa2c9041f9708a2880ffabc45c58c2045e13de90200e74f182092c7d9ed","sourceUrl":"https://arxiv.org/abs/2610.39020","contentSha256":"dc8bbb425111e3b414ffe63f69edcaf0257ca053ac60d2cbb1d8d94f2a4e6b59","textHash":"7d0962984b7e69e4c87e6ca815ffb254ae30c6566ba55c1ff1dc4758024ece5c","observedAt":"2026-11-11T08:56:07.537Z"}}],"participationCoverage":{"total":6,"returned":6,"complete":true,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":null,"sourceUrl":"https://arxiv.org/abs/2610.39020","snapshotId":"29017aa2c9041f9708a2880ffabc45c58c2045e13de90200e74f182092c7d9ed","contentSha256":"dc8bbb425111e3b414ffe63f69edcaf0257ca053ac60d2cbb1d8d94f2a4e6b59","textHash":"7d0962984b7e69e4c87e6ca815ffb254ae30c6566ba55c1ff1dc4758024ece5c","observedAt":"2026-11-11T08:56:07.537Z","basis":"publisher-page-metadata","sourceLinks":46,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/e2babda51d94af28b41b581a7d73079d9cc83c7e82c3ef7dda44dc18e661bfd0/source"}},"canonicalUrl":"https://escapemeridian.com/readings/the-paper-argues-opd-learns-where-reasoning-paths-stay-open-e2babda5","formats":{"html":"/readings/the-paper-argues-opd-learns-where-reasoning-paths-stay-open-e2babda5","markdown":"/readings/the-paper-argues-opd-learns-where-reasoning-paths-stay-open-e2babda5.md","json":"/readings/the-paper-argues-opd-learns-where-reasoning-paths-stay-open-e2babda5.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"},{"id":"the-sequence","name":"The Sequence","kind":"organization","context":"The Sequence publishes machine-learning analysis and interviews whose guest answers retain their own attribution.","referenceUrls":["https://thesequence.substack.com/","https://thesequence.substack.com/p/the-sequence-chat-emad-mostaque-stability"],"path":"/entities/the-sequence","json":"/entities/the-sequence.json","markdown":"/entities/the-sequence.md"}]},"researchQuestions":[],"withdrawn":false,"references":{"schemaVersion":1,"target":{"type":"reading","id":"e2babda51d94af28b41b581a7d73079d9cc83c7e82c3ef7dda44dc18e661bfd0","revision":"daadfca5-7868-4fbd-8892-e560800eca6c","revisionKind":"publication-id","title":"The paper argues OPD learns where reasoning paths stay open","url":"https://escapemeridian.com/readings/the-paper-argues-opd-learns-where-reasoning-paths-stay-open-e2babda5"},"source":{"type":"source","id":"https://arxiv.org/abs/2610.39020","revision":"7d0962984b7e69e4c87e6ca815ffb254ae30c6566ba55c1ff1dc4758024ece5c","revisionKind":"sha256-source-text","title":"Distilling the Shape of Uncertainty: What On-Policy Distillation Learns and How to Predict It","url":"https://arxiv.org/abs/2610.39020"},"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":"e2babda51d94af28b41b581a7d73079d9cc83c7e82c3ef7dda44dc18e661bfd0","title":"The paper argues OPD learns where reasoning paths stay open","argument":"Authors Shiping Yang, Chuanqi Tang, Hossein Hajimirsadeghi, Layla El Asri, Mohammad Amin Shabani, and Angel X. Chang argue that on-policy distillation (OPD) learns more than which candidate tokens a student considers. It also learns where the student keeps reasoning paths open, which the authors call the “shape” of uncertainty.\n\nThe authors introduce entropy shape—the sequence of next-token entropies—and Entropy Shape Correlation (ESC), a measure of student-teacher agreement on that sequence. They report that intervening on a teacher’s entropy while preserving its token preferences indicates that where to branch causally drives OPD gains, independently of how to branch.\n\nThey also frame predicting OPD outcomes as a teacher-selection task. In their evaluation protocol, pre-OPD ESC selects the best teacher for both students and outperforms existing predictors. If this measure helps identify a useful teacher before distillation, it could make teacher selection more informed; the abstract does not establish how well the result transfers beyond that protocol. What results would show whether pre-OPD ESC predicts teacher quality for students and settings outside the authors’ evaluation protocol?","topicIds":["intelligence-and-discovery"],"domain":"science","claimType":"attributed reading","limits":"The evidence is the paper’s abstract; it does not provide the evaluation protocol’s details, measurements, or the scope of students and settings tested. The abstract does not establish whether pre-OPD ESC predicts teacher quality beyond the authors’ evaluation protocol.","speaker":{"name":"Yang, Shiping, Tang, Chuanqi, Hajimirsadeghi, Hossein, Asri, Layla El, Shabani, Mohammad Amin, Chang, Angel X.","roleAtSource":"Article author"},"source":{"title":"Distilling the Shape of Uncertainty: What On-Policy Distillation Learns and How to Predict It","publisher":"arXiv cs.AI","format":"Article","url":"https://arxiv.org/abs/2610.39020","publishedAt":"2026-10-28","dateBasis":"Date basis not provided."},"evidence":[{"text":"The authors argue that OPD learns both which candidate tokens to consider and where to keep reasoning paths open.","excerpt":"OPD learns not only \\textit{how to branch}, i.e., which candidate tokens to consider, but also \\textit{where to branch}, i.e., where to keep reasoning paths open.","url":"https://arxiv.org/abs/2610.39020","relationship":"Source-stated"},{"text":"The authors introduce entropy shape and ESC to characterize and measure agreement on where reasoning paths remain open.","excerpt":"We introduce the entropy shape, the sequence of next-token entropies, to characterize the latter, and propose the Entropy Shape Correlation (ESC) to measure student-teacher agreement in entropy shape.","url":"https://arxiv.org/abs/2610.39020","relationship":"Source-stated"},{"text":"The authors report that where to branch causally drives OPD gains independently of how to branch, based on tracking ESC and intervening on teacher entropy while preserving token preferences.","excerpt":"intervening on a teacher's entropy while preserving its token preferences, we empirically show that \\textit{where to branch} causally drives OPD gains, independently of \\textit{how to branch}.","url":"https://arxiv.org/abs/2610.39020","relationship":"Source-stated"},{"text":"Under the authors’ evaluation protocol, pre-OPD ESC selects the best teacher for both students and outperforms existing predictors.","excerpt":"under which pre-OPD ESC selects the best teacher for both students and outperforms existing predictors.","url":"https://arxiv.org/abs/2610.39020","relationship":"Source-stated"}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2610.39020","checkedAt":"2026-11-11T08:56:07.537Z","items":[],"note":"No suitable media was found by the supported page extraction."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"What results would show whether pre-OPD ESC predicts teacher quality for students and settings outside the authors’ evaluation protocol?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"The paper argues OPD learns where reasoning paths stay open","summary":"Authors Shiping Yang, Chuanqi Tang, Hossein Hajimirsadeghi, Layla El Asri, Mohammad Amin Shabani, and Angel X. Chang argue that on-policy distillation (OPD) learns more than which candidate tokens a student considers. It also learns where the student keeps reasoning paths open, which the authors call the “shape” of uncertainty.\n\nThe authors introduce entropy shape—the sequence of next-token entropies—and Entropy Shape Correlation (ESC), a measure of student-teacher agreement on that sequence. They report that intervening on a teacher’s entropy while preserving its token preferences indicates that where to branch causally drives OPD gains, independently of how to branch.\n\nThey also frame predicting OPD outcomes as a teacher-selection task. In their evaluation protocol, pre-OPD ESC selects the best teacher for both students and outperforms existing predictors. If this measure helps identify a useful teacher before distillation, it could make teacher selection more informed; the abstract does not establish how well the result transfers beyond that protocol. What results would show whether pre-OPD ESC predicts teacher quality for students and settings outside the authors’ evaluation protocol?","question":"What results would show whether pre-OPD ESC predicts teacher quality for students and settings outside the authors’ evaluation protocol?","claims":[{"text":"The authors argue that OPD learns both which candidate tokens to consider and where to keep reasoning paths open.","excerpt":"OPD learns not only \\textit{how to branch}, i.e., which candidate tokens to consider, but also \\textit{where to branch}, i.e., where to keep reasoning paths open.","support":"source-stated"},{"text":"The authors introduce entropy shape and ESC to characterize and measure agreement on where reasoning paths remain open.","excerpt":"We introduce the entropy shape, the sequence of next-token entropies, to characterize the latter, and propose the Entropy Shape Correlation (ESC) to measure student-teacher agreement in entropy shape.","support":"source-stated"},{"text":"The authors report that where to branch causally drives OPD gains independently of how to branch, based on tracking ESC and intervening on teacher entropy while preserving token preferences.","excerpt":"intervening on a teacher's entropy while preserving its token preferences, we empirically show that \\textit{where to branch} causally drives OPD gains, independently of \\textit{how to branch}.","support":"source-stated"},{"text":"Under the authors’ evaluation protocol, pre-OPD ESC selects the best teacher for both students and outperforms existing predictors.","excerpt":"under which pre-OPD ESC selects the best teacher for both students and outperforms existing predictors.","support":"source-stated"}],"limitations":["The evidence is the paper’s abstract; it does not provide the evaluation protocol’s details, measurements, or the scope of students and settings tested.","The abstract does not establish whether pre-OPD ESC predicts teacher quality beyond the authors’ evaluation protocol."],"mediaId":"f4a7a7198e1aa606bea8a43ac56f9a1c999a99327ce8148192af7551aacc86d6","correctionReason":null,"id":"e2babda51d94af28b41b581a7d73079d9cc83c7e82c3ef7dda44dc18e661bfd0","publicationId":"daadfca5-7868-4fbd-8892-e560800eca6c","version":1,"sourceUrl":"https://arxiv.org/abs/2610.39020","sourceTitle":"Distilling the Shape of Uncertainty: What On-Policy Distillation Learns and How to Predict It","publisher":"arXiv cs.AI","publishedAt":null,"originalPublished":{"value":"2026-10-28","precision":"day","basis":"publisher_citation_date"},"firstObservedAt":"2026-10-30T06:42:44.412Z","createdAt":"2026-11-11T08:25:25.106Z","historical":true,"entityIds":["arxiv"],"topicIds":["intelligence-and-discovery"],"media":[{"id":"f4a7a7198e1aa606bea8a43ac56f9a1c999a99327ce8148192af7551aacc86d6","kind":"article","url":"https://arxiv.org/abs/2610.39020","sourceUrl":"https://arxiv.org/abs/2610.39020","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-10-31T11:46:50.525Z","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/2610.39020","checkedAt":"2026-11-11T08:56:07.537Z","items":[],"note":"No suitable media was found by the supported page extraction."},"editorialDecision":{"outcome":"accept","reasonCodes":["material_development"],"rubricVersion":"stream-admission-v5","mediaId":"f4a7a7198e1aa606bea8a43ac56f9a1c999a99327ce8148192af7551aacc86d6","confidence":0.8400000000000001,"provider":"typesafe","model":"jev-1.13.0","inputHash":"65a29790ce85216f1ed79c8d1ca188c384e454af5f4a9dc7b89e5ae7555944f8","decidedAt":"2026-10-31T11:46:52.237Z"},"provenance":{"textHash":"7d0962984b7e69e4c87e6ca815ffb254ae30c6566ba55c1ff1dc4758024ece5c","observationId":"4858c3e836a9c21eb1834f6ebced66f1b13a4ed4b8497895227608cc71448934","retrievedAt":"2026-10-31T11:46:50.280Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"5bd6b3ff-0559-44da-83d4-bee6193ab45f"},"reviewer":{"model":"gpt-6-luna","runId":"1a7e2cb5-e4b6-41f7-8112-9a6ceb2dbb78"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-11T08:25:25.106Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Yang, Shiping","url":null,"kind":"unknown"},{"name":"Tang, Chuanqi","url":null,"kind":"unknown"},{"name":"Hajimirsadeghi, Hossein","url":null,"kind":"unknown"},{"name":"Asri, Layla El","url":null,"kind":"unknown"},{"name":"Shabani, Mohammad Amin","url":null,"kind":"unknown"},{"name":"Chang, Angel X.","url":null,"kind":"unknown"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Yang, Shiping","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Yang, Shiping","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.39020","locator":"head/meta[name=author|citation_author]"},"assertionId":"29017aa2c9041f9708a2880ffabc45c58c2045e13de90200e74f182092c7d9ed:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-11T08:56:07.537Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"29017aa2c9041f9708a2880ffabc45c58c2045e13de90200e74f182092c7d9ed","sourceUrl":"https://arxiv.org/abs/2610.39020","contentSha256":"dc8bbb425111e3b414ffe63f69edcaf0257ca053ac60d2cbb1d8d94f2a4e6b59","textHash":"7d0962984b7e69e4c87e6ca815ffb254ae30c6566ba55c1ff1dc4758024ece5c","observedAt":"2026-11-11T08:56:07.537Z"}},{"actor":{"name":"Tang, Chuanqi","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Tang, Chuanqi","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.39020","locator":"head/meta[name=author|citation_author]"},"assertionId":"29017aa2c9041f9708a2880ffabc45c58c2045e13de90200e74f182092c7d9ed:participation:1","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-11T08:56:07.537Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"29017aa2c9041f9708a2880ffabc45c58c2045e13de90200e74f182092c7d9ed","sourceUrl":"https://arxiv.org/abs/2610.39020","contentSha256":"dc8bbb425111e3b414ffe63f69edcaf0257ca053ac60d2cbb1d8d94f2a4e6b59","textHash":"7d0962984b7e69e4c87e6ca815ffb254ae30c6566ba55c1ff1dc4758024ece5c","observedAt":"2026-11-11T08:56:07.537Z"}},{"actor":{"name":"Hajimirsadeghi, Hossein","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Hajimirsadeghi, Hossein","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.39020","locator":"head/meta[name=author|citation_author]"},"assertionId":"29017aa2c9041f9708a2880ffabc45c58c2045e13de90200e74f182092c7d9ed:participation:2","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-11T08:56:07.537Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"29017aa2c9041f9708a2880ffabc45c58c2045e13de90200e74f182092c7d9ed","sourceUrl":"https://arxiv.org/abs/2610.39020","contentSha256":"dc8bbb425111e3b414ffe63f69edcaf0257ca053ac60d2cbb1d8d94f2a4e6b59","textHash":"7d0962984b7e69e4c87e6ca815ffb254ae30c6566ba55c1ff1dc4758024ece5c","observedAt":"2026-11-11T08:56:07.537Z"}},{"actor":{"name":"Asri, Layla El","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Asri, Layla El","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.39020","locator":"head/meta[name=author|citation_author]"},"assertionId":"29017aa2c9041f9708a2880ffabc45c58c2045e13de90200e74f182092c7d9ed:participation:3","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-11T08:56:07.537Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"29017aa2c9041f9708a2880ffabc45c58c2045e13de90200e74f182092c7d9ed","sourceUrl":"https://arxiv.org/abs/2610.39020","contentSha256":"dc8bbb425111e3b414ffe63f69edcaf0257ca053ac60d2cbb1d8d94f2a4e6b59","textHash":"7d0962984b7e69e4c87e6ca815ffb254ae30c6566ba55c1ff1dc4758024ece5c","observedAt":"2026-11-11T08:56:07.537Z"}},{"actor":{"name":"Shabani, Mohammad Amin","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Shabani, Mohammad Amin","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.39020","locator":"head/meta[name=author|citation_author]"},"assertionId":"29017aa2c9041f9708a2880ffabc45c58c2045e13de90200e74f182092c7d9ed:participation:4","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-11T08:56:07.537Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"29017aa2c9041f9708a2880ffabc45c58c2045e13de90200e74f182092c7d9ed","sourceUrl":"https://arxiv.org/abs/2610.39020","contentSha256":"dc8bbb425111e3b414ffe63f69edcaf0257ca053ac60d2cbb1d8d94f2a4e6b59","textHash":"7d0962984b7e69e4c87e6ca815ffb254ae30c6566ba55c1ff1dc4758024ece5c","observedAt":"2026-11-11T08:56:07.537Z"}},{"actor":{"name":"Chang, Angel X.","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Chang, Angel X.","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.39020","locator":"head/meta[name=author|citation_author]"},"assertionId":"29017aa2c9041f9708a2880ffabc45c58c2045e13de90200e74f182092c7d9ed:participation:5","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-11T08:56:07.537Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"29017aa2c9041f9708a2880ffabc45c58c2045e13de90200e74f182092c7d9ed","sourceUrl":"https://arxiv.org/abs/2610.39020","contentSha256":"dc8bbb425111e3b414ffe63f69edcaf0257ca053ac60d2cbb1d8d94f2a4e6b59","textHash":"7d0962984b7e69e4c87e6ca815ffb254ae30c6566ba55c1ff1dc4758024ece5c","observedAt":"2026-11-11T08:56:07.537Z"}}],"participationCoverage":{"total":6,"returned":6,"complete":true,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":null,"sourceUrl":"https://arxiv.org/abs/2610.39020","snapshotId":"29017aa2c9041f9708a2880ffabc45c58c2045e13de90200e74f182092c7d9ed","contentSha256":"dc8bbb425111e3b414ffe63f69edcaf0257ca053ac60d2cbb1d8d94f2a4e6b59","textHash":"7d0962984b7e69e4c87e6ca815ffb254ae30c6566ba55c1ff1dc4758024ece5c","observedAt":"2026-11-11T08:56:07.537Z","basis":"publisher-page-metadata","sourceLinks":46,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/e2babda51d94af28b41b581a7d73079d9cc83c7e82c3ef7dda44dc18e661bfd0/source"}}}],"audits":[{"publicationId":"daadfca5-7868-4fbd-8892-e560800eca6c","verdict":"uncertain","findings":[{"dimension":"attribution","status":"pass","explanation":"The six listed authors match the arXiv page. The claims are framed as what the authors argue, introduce or report, not as established fact or EM opinion.","excerpt":"Authors:Shiping Yang, Chuanqi Tang, Hossein Hajimirsadeghi, Layla El Asri, Mohammad Amin Shabani, Angel X. Chang"},{"dimension":"summary","status":"uncertain","explanation":"The reading tracks the abstract faithfully, but a non-specialist may not follow it. It never says that OPD is a post-training method in which a student language model learns from a teacher model. It also leaves 'tokens' and 'next-token entropies' unexplained (add tooltips). It omits the motivating puzzle that a stronger teacher can fail to improve the student. 'Both students' appears without saying two student models were tested. Smallest repair: in the first paragraph, add the post-training context and the stronger-teacher puzzle. Tooltips should cover OPD, token and entropy.","excerpt":"On-policy distillation (OPD) is increasingly adopted in the post-training of large language models. However, a stronger teacher may fail to improve the student."},{"dimension":"date","status":"pass","explanation":"originalPublished 2026-10-28 matches the v1 submission date. The later feed date (Oct 31) was correctly not used.","excerpt":"[Submitted on 28 Oct 2026]"},{"dimension":"support","status":"pass","explanation":"Each claim is backed by an exact abstract excerpt. The causal finding is attributed as the authors' empirical claim. The teacher-selection result is scoped to their evaluation protocol. The benefit statement is explicitly conditional, and the limitations note that the evidence is abstract-only. The question follows from that unresolved scope.","excerpt":"we formulate OPD outcome prediction as a teacher selection task and establish an evaluation protocol for it, under which pre-OPD ESC selects the best teacher for both students and outperforms existing predictors."},{"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":"2026-11-11T08:25:25.106Z","model":"anthropic/claude-opus-5.5"}],"correctionUrl":"https://escapemeridian.com/corrections/"}