{"id":"b17157a7f4f43f7dde53453c3116a873f032f6d2edeac3ae07b9371b32f5733b","title":"Retrieving from Your Own Mistakes: On-Policy Training of Embedding Models for Multi-Hop Retrieval-Augmented Generation","argument":"In their paper, Chenjie Qiu and coauthors argue that a retriever for multi-hop retrieval-augmented generation should train on the retrieval states it reaches itself, including states containing mistakes. Chain embedding retrievers find evidence step by step, updating the question representation with passages already retrieved. The authors say training only on gold histories leaves a mismatch: at test time, a retriever conditions on its own passages, and by the third step about half or more of its states contain a wrong passage. Their on-policy state training approach labels each visited state with the earliest missing gold passage, masks other missing passages, and trains on these states alongside teacher-forced ones. The authors report that this retrieves complete evidence more often than teacher-forced training and compared systems, transfers without further training to harder long-chain variants of HotpotQA and 2WikiMultiHopQA, and recovers from its own mistakes more often on the same states. The idea matters because errors early in a multi-step search can make later evidence harder to find; the paper tests whether practicing on those error-shaped states helps the retriever recover. The abstract does not give the size of the gains or enough detail to assess their practical significance. How large are the reported gains, and how do they vary across the evaluated datasets and.","topicIds":["intelligence-and-discovery"],"domain":"science","claimType":"attributed reading","limits":"The supplied evidence is the paper’s abstract, which does not state the size of the reported gains or provide detailed results by dataset or comparison system. The abstract describes harder long-chain variants of HotpotQA and 2WikiMultiHopQA that the authors construct; the supplied evidence does not describe those variants in detail.","speaker":{"name":"Qiu, Chenjie, Zeng, Zixuan, Shi, Shaobin, Chen, Bingfeng, Luo, Dongsheng, Xu, Boyan","roleAtSource":"Article author"},"source":{"title":"Retrieving from Your Own Mistakes: On-Policy Training of Embedding Models for Multi-Hop Retrieval-Augmented Generation","publisher":"arXiv cs.AI","format":"Article","url":"https://arxiv.org/abs/2610.40409","publishedAt":"2026-10-29","dateBasis":"Date basis not provided."},"evidence":[{"text":"The authors say a retriever trained only on gold histories can face a mismatch at test time, when it conditions on its own retrieved passages, which may be wrong.","excerpt":"At test time the retriever conditions on its own retrievals, which soon contain mistakes","url":"https://arxiv.org/abs/2610.40409","relationship":"Source-stated"},{"text":"The authors report that by the third step, about half or more of the states visited contain a wrong passage, and that these passages reduce the chance of retrieving missing gold evidence.","excerpt":"by the third step, about half or more of the states it visits hold a wrong passage, and these passages lower its chance of retrieving a missing gold passage","url":"https://arxiv.org/abs/2610.40409","relationship":"Source-stated"},{"text":"The proposed on-policy state training trains the encoder using states reached by its current retrievals, including states built from its own mistakes.","excerpt":"We propose on-policy state training, which trains the encoder on the states it reaches itself.","url":"https://arxiv.org/abs/2610.40409","relationship":"Source-stated"},{"text":"The authors report that the resulting retriever retrieves complete evidence more often than teacher-forced training and the systems they compare.","excerpt":"The resulting retriever retrieves complete evidence more often than teacher-forced training and the systems we compare","url":"https://arxiv.org/abs/2610.40409","relationship":"Source-stated"},{"text":"The authors report transfer without further training to harder long-chain variants of HotpotQA and 2WikiMultiHopQA that they construct.","excerpt":"it transfers without further training to harder long-chain variants of HotpotQA and 2WikiMultiHopQA that we construct","url":"https://arxiv.org/abs/2610.40409","relationship":"Source-stated"},{"text":"The paper’s approach targets the problem that a retriever’s own earlier mistakes can make later missing evidence harder to retrieve.","excerpt":"these passages lower its chance of retrieving a missing gold passage well beyond what the difficulty of the states explains","url":"https://arxiv.org/abs/2610.40409","relationship":"Escape Meridian interpretation"}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2610.40409","checkedAt":"2026-11-13T09:10:07.394Z","items":[],"note":"No suitable media was found by the supported page extraction."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"How large are the reported gains, and how do they vary across the evaluated datasets and systems?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"Retrieving from Your Own Mistakes: On-Policy Training of Embedding Models for Multi-Hop Retrieval-Augmented Generation","summary":"In their paper, Chenjie Qiu and coauthors argue that a retriever for multi-hop retrieval-augmented generation should train on the retrieval states it reaches itself, including states containing mistakes. Chain embedding retrievers find evidence step by step, updating the question representation with passages already retrieved. The authors say training only on gold histories leaves a mismatch: at test time, a retriever conditions on its own passages, and by the third step about half or more of its states contain a wrong passage. Their on-policy state training approach labels each visited state with the earliest missing gold passage, masks other missing passages, and trains on these states alongside teacher-forced ones. The authors report that this retrieves complete evidence more often than teacher-forced training and compared systems, transfers without further training to harder long-chain variants of HotpotQA and 2WikiMultiHopQA, and recovers from its own mistakes more often on the same states. The idea matters because errors early in a multi-step search can make later evidence harder to find; the paper tests whether practicing on those error-shaped states helps the retriever recover. The abstract does not give the size of the gains or enough detail to assess their practical significance. How large are the reported gains, and how do they vary across the evaluated datasets and.","question":"How large are the reported gains, and how do they vary across the evaluated datasets and systems?","claims":[{"text":"The authors say a retriever trained only on gold histories can face a mismatch at test time, when it conditions on its own retrieved passages, which may be wrong.","excerpt":"At test time the retriever conditions on its own retrievals, which soon contain mistakes","support":"source-stated"},{"text":"The authors report that by the third step, about half or more of the states visited contain a wrong passage, and that these passages reduce the chance of retrieving missing gold evidence.","excerpt":"by the third step, about half or more of the states it visits hold a wrong passage, and these passages lower its chance of retrieving a missing gold passage","support":"source-stated"},{"text":"The proposed on-policy state training trains the encoder using states reached by its current retrievals, including states built from its own mistakes.","excerpt":"We propose on-policy state training, which trains the encoder on the states it reaches itself.","support":"source-stated"},{"text":"The authors report that the resulting retriever retrieves complete evidence more often than teacher-forced training and the systems they compare.","excerpt":"The resulting retriever retrieves complete evidence more often than teacher-forced training and the systems we compare","support":"source-stated"},{"text":"The authors report transfer without further training to harder long-chain variants of HotpotQA and 2WikiMultiHopQA that they construct.","excerpt":"it transfers without further training to harder long-chain variants of HotpotQA and 2WikiMultiHopQA that we construct","support":"source-stated"},{"text":"The paper’s approach targets the problem that a retriever’s own earlier mistakes can make later missing evidence harder to retrieve.","excerpt":"these passages lower its chance of retrieving a missing gold passage well beyond what the difficulty of the states explains","support":"interpretation"}],"limitations":["The supplied evidence is the paper’s abstract, which does not state the size of the reported gains or provide detailed results by dataset or comparison system.","The abstract describes harder long-chain variants of HotpotQA and 2WikiMultiHopQA that the authors construct; the supplied evidence does not describe those variants in detail."],"mediaId":"1c020a99b0a958ac24df25bff425fc078015c8ca6a9c4453f3e4cc6bc81ef9b2","correctionReason":null,"id":"b17157a7f4f43f7dde53453c3116a873f032f6d2edeac3ae07b9371b32f5733b","publicationId":"d08256f4-99a6-41f9-817f-b6ef0a597c9c","version":1,"sourceUrl":"https://arxiv.org/abs/2610.40409","sourceTitle":"Retrieving from Your Own Mistakes: On-Policy Training of Embedding Models for Multi-Hop Retrieval-Augmented Generation","publisher":"arXiv cs.AI","publishedAt":null,"originalPublished":{"value":"2026-10-29","precision":"day","basis":"publisher_citation_date"},"firstObservedAt":"2026-10-30T06:42:53.592Z","createdAt":"2026-11-13T07:03:07.331Z","historical":true,"entityIds":["arxiv"],"topicIds":["intelligence-and-discovery"],"media":[{"id":"1c020a99b0a958ac24df25bff425fc078015c8ca6a9c4453f3e4cc6bc81ef9b2","kind":"article","url":"https://arxiv.org/abs/2610.40409","sourceUrl":"https://arxiv.org/abs/2610.40409","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-10-31T09:31:07.096Z","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/2610.40409","checkedAt":"2026-11-13T09:10:07.394Z","items":[],"note":"No suitable media was found by the supported page extraction."},"editorialDecision":{"outcome":"accept","reasonCodes":["material_development"],"rubricVersion":"stream-admission-v5","mediaId":"1c020a99b0a958ac24df25bff425fc078015c8ca6a9c4453f3e4cc6bc81ef9b2","confidence":0.8200000000000001,"provider":"typesafe","model":"jev-1.13.0","inputHash":"4f8f5d529ca5b7183973fd94914ecdb2578baec2efb7b3f9908e106d1f97e61b","decidedAt":"2026-10-31T09:31:07.910Z"},"provenance":{"textHash":"c4b521a32436de807e84e36c0822b4ffd39efd14d51f2f3e33f7aef0faa335b4","observationId":"8bdfd9e8beeb32c2897f4e18ecd0e7e6a0af4b3557f428dc243ebfae7f8bb900","retrievedAt":"2026-10-31T09:31:07.007Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"428f50fe-2f2a-4708-8591-804a37db19ee"},"reviewer":{"model":"gpt-6-luna","runId":"f541e54d-6043-4703-8fb6-3d265597c7e9"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-13T07:03:07.331Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Qiu, Chenjie","url":null,"kind":"unknown"},{"name":"Zeng, Zixuan","url":null,"kind":"unknown"},{"name":"Shi, Shaobin","url":null,"kind":"unknown"},{"name":"Chen, Bingfeng","url":null,"kind":"unknown"},{"name":"Luo, Dongsheng","url":null,"kind":"unknown"},{"name":"Xu, Boyan","url":null,"kind":"unknown"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Qiu, Chenjie","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Qiu, Chenjie","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.40409","locator":"head/meta[name=author|citation_author]"},"assertionId":"b6a230068294134278e63603d3d8b19af6c1d0a47403ed347d46a1d1158a191c:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-13T09:10:07.394Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"b6a230068294134278e63603d3d8b19af6c1d0a47403ed347d46a1d1158a191c","sourceUrl":"https://arxiv.org/abs/2610.40409","contentSha256":"17a8ca064e4e27d1c3169945ee09ae8d5f7c7f64e10dc833acd6f31b6db24dc5","textHash":"c4b521a32436de807e84e36c0822b4ffd39efd14d51f2f3e33f7aef0faa335b4","observedAt":"2026-11-13T09:10:07.394Z"}},{"actor":{"name":"Zeng, Zixuan","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Zeng, Zixuan","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.40409","locator":"head/meta[name=author|citation_author]"},"assertionId":"b6a230068294134278e63603d3d8b19af6c1d0a47403ed347d46a1d1158a191c:participation:1","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-13T09:10:07.394Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"b6a230068294134278e63603d3d8b19af6c1d0a47403ed347d46a1d1158a191c","sourceUrl":"https://arxiv.org/abs/2610.40409","contentSha256":"17a8ca064e4e27d1c3169945ee09ae8d5f7c7f64e10dc833acd6f31b6db24dc5","textHash":"c4b521a32436de807e84e36c0822b4ffd39efd14d51f2f3e33f7aef0faa335b4","observedAt":"2026-11-13T09:10:07.394Z"}},{"actor":{"name":"Shi, Shaobin","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Shi, Shaobin","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.40409","locator":"head/meta[name=author|citation_author]"},"assertionId":"b6a230068294134278e63603d3d8b19af6c1d0a47403ed347d46a1d1158a191c:participation:2","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-13T09:10:07.394Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"b6a230068294134278e63603d3d8b19af6c1d0a47403ed347d46a1d1158a191c","sourceUrl":"https://arxiv.org/abs/2610.40409","contentSha256":"17a8ca064e4e27d1c3169945ee09ae8d5f7c7f64e10dc833acd6f31b6db24dc5","textHash":"c4b521a32436de807e84e36c0822b4ffd39efd14d51f2f3e33f7aef0faa335b4","observedAt":"2026-11-13T09:10:07.394Z"}},{"actor":{"name":"Chen, Bingfeng","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Chen, Bingfeng","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.40409","locator":"head/meta[name=author|citation_author]"},"assertionId":"b6a230068294134278e63603d3d8b19af6c1d0a47403ed347d46a1d1158a191c:participation:3","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-13T09:10:07.394Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"b6a230068294134278e63603d3d8b19af6c1d0a47403ed347d46a1d1158a191c","sourceUrl":"https://arxiv.org/abs/2610.40409","contentSha256":"17a8ca064e4e27d1c3169945ee09ae8d5f7c7f64e10dc833acd6f31b6db24dc5","textHash":"c4b521a32436de807e84e36c0822b4ffd39efd14d51f2f3e33f7aef0faa335b4","observedAt":"2026-11-13T09:10:07.394Z"}},{"actor":{"name":"Luo, Dongsheng","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Luo, Dongsheng","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.40409","locator":"head/meta[name=author|citation_author]"},"assertionId":"b6a230068294134278e63603d3d8b19af6c1d0a47403ed347d46a1d1158a191c:participation:4","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-13T09:10:07.394Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"b6a230068294134278e63603d3d8b19af6c1d0a47403ed347d46a1d1158a191c","sourceUrl":"https://arxiv.org/abs/2610.40409","contentSha256":"17a8ca064e4e27d1c3169945ee09ae8d5f7c7f64e10dc833acd6f31b6db24dc5","textHash":"c4b521a32436de807e84e36c0822b4ffd39efd14d51f2f3e33f7aef0faa335b4","observedAt":"2026-11-13T09:10:07.394Z"}},{"actor":{"name":"Xu, Boyan","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Xu, Boyan","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.40409","locator":"head/meta[name=author|citation_author]"},"assertionId":"b6a230068294134278e63603d3d8b19af6c1d0a47403ed347d46a1d1158a191c:participation:5","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-13T09:10:07.394Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"b6a230068294134278e63603d3d8b19af6c1d0a47403ed347d46a1d1158a191c","sourceUrl":"https://arxiv.org/abs/2610.40409","contentSha256":"17a8ca064e4e27d1c3169945ee09ae8d5f7c7f64e10dc833acd6f31b6db24dc5","textHash":"c4b521a32436de807e84e36c0822b4ffd39efd14d51f2f3e33f7aef0faa335b4","observedAt":"2026-11-13T09:10:07.394Z"}}],"participationCoverage":{"total":6,"returned":6,"complete":true,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":null,"sourceUrl":"https://arxiv.org/abs/2610.40409","snapshotId":"b6a230068294134278e63603d3d8b19af6c1d0a47403ed347d46a1d1158a191c","contentSha256":"17a8ca064e4e27d1c3169945ee09ae8d5f7c7f64e10dc833acd6f31b6db24dc5","textHash":"c4b521a32436de807e84e36c0822b4ffd39efd14d51f2f3e33f7aef0faa335b4","observedAt":"2026-11-13T09:10:07.394Z","basis":"publisher-page-metadata","sourceLinks":44,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/b17157a7f4f43f7dde53453c3116a873f032f6d2edeac3ae07b9371b32f5733b/source"}},"canonicalUrl":"https://escapemeridian.com/readings/retrieving-from-your-own-mistakes-on-policy-training-of-b17157a7","formats":{"html":"/readings/retrieving-from-your-own-mistakes-on-policy-training-of-b17157a7","markdown":"/readings/retrieving-from-your-own-mistakes-on-policy-training-of-b17157a7.md","json":"/readings/retrieving-from-your-own-mistakes-on-policy-training-of-b17157a7.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-rag","name":"Retrieval-augmented generation","explanation":"A setup in which a model first looks up relevant documents and then answers using them.","referenceUrl":"https://developers.google.com/machine-learning/glossary#retrieval-augmented-generation-rag"}],"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":"time","name":"TIME","kind":"organization","context":"TIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.","referenceUrls":["https://time.com/"],"path":"/entities/time","json":"/entities/time.json","markdown":"/entities/time.md"}]},"researchQuestions":[],"withdrawn":false,"references":{"schemaVersion":1,"target":{"type":"reading","id":"b17157a7f4f43f7dde53453c3116a873f032f6d2edeac3ae07b9371b32f5733b","revision":"d08256f4-99a6-41f9-817f-b6ef0a597c9c","revisionKind":"publication-id","title":"Retrieving from Your Own Mistakes: On-Policy Training of Embedding Models for Multi-Hop Retrieval-Augmented Generation","url":"https://escapemeridian.com/readings/retrieving-from-your-own-mistakes-on-policy-training-of-b17157a7"},"source":{"type":"source","id":"https://arxiv.org/abs/2610.40409","revision":"c4b521a32436de807e84e36c0822b4ffd39efd14d51f2f3e33f7aef0faa335b4","revisionKind":"sha256-source-text","title":"Retrieving from Your Own Mistakes: On-Policy Training of Embedding Models for Multi-Hop Retrieval-Augmented Generation","url":"https://arxiv.org/abs/2610.40409"},"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":"b17157a7f4f43f7dde53453c3116a873f032f6d2edeac3ae07b9371b32f5733b","title":"Retrieving from Your Own Mistakes: On-Policy Training of Embedding Models for Multi-Hop Retrieval-Augmented Generation","argument":"In their paper, Chenjie Qiu and coauthors argue that a retriever for multi-hop retrieval-augmented generation should train on the retrieval states it reaches itself, including states containing mistakes. Chain embedding retrievers find evidence step by step, updating the question representation with passages already retrieved. The authors say training only on gold histories leaves a mismatch: at test time, a retriever conditions on its own passages, and by the third step about half or more of its states contain a wrong passage. Their on-policy state training approach labels each visited state with the earliest missing gold passage, masks other missing passages, and trains on these states alongside teacher-forced ones. The authors report that this retrieves complete evidence more often than teacher-forced training and compared systems, transfers without further training to harder long-chain variants of HotpotQA and 2WikiMultiHopQA, and recovers from its own mistakes more often on the same states. The idea matters because errors early in a multi-step search can make later evidence harder to find; the paper tests whether practicing on those error-shaped states helps the retriever recover. The abstract does not give the size of the gains or enough detail to assess their practical significance. How large are the reported gains, and how do they vary across the evaluated datasets and.","topicIds":["intelligence-and-discovery"],"domain":"science","claimType":"attributed reading","limits":"The supplied evidence is the paper’s abstract, which does not state the size of the reported gains or provide detailed results by dataset or comparison system. The abstract describes harder long-chain variants of HotpotQA and 2WikiMultiHopQA that the authors construct; the supplied evidence does not describe those variants in detail.","speaker":{"name":"Qiu, Chenjie, Zeng, Zixuan, Shi, Shaobin, Chen, Bingfeng, Luo, Dongsheng, Xu, Boyan","roleAtSource":"Article author"},"source":{"title":"Retrieving from Your Own Mistakes: On-Policy Training of Embedding Models for Multi-Hop Retrieval-Augmented Generation","publisher":"arXiv cs.AI","format":"Article","url":"https://arxiv.org/abs/2610.40409","publishedAt":"2026-10-29","dateBasis":"Date basis not provided."},"evidence":[{"text":"The authors say a retriever trained only on gold histories can face a mismatch at test time, when it conditions on its own retrieved passages, which may be wrong.","excerpt":"At test time the retriever conditions on its own retrievals, which soon contain mistakes","url":"https://arxiv.org/abs/2610.40409","relationship":"Source-stated"},{"text":"The authors report that by the third step, about half or more of the states visited contain a wrong passage, and that these passages reduce the chance of retrieving missing gold evidence.","excerpt":"by the third step, about half or more of the states it visits hold a wrong passage, and these passages lower its chance of retrieving a missing gold passage","url":"https://arxiv.org/abs/2610.40409","relationship":"Source-stated"},{"text":"The proposed on-policy state training trains the encoder using states reached by its current retrievals, including states built from its own mistakes.","excerpt":"We propose on-policy state training, which trains the encoder on the states it reaches itself.","url":"https://arxiv.org/abs/2610.40409","relationship":"Source-stated"},{"text":"The authors report that the resulting retriever retrieves complete evidence more often than teacher-forced training and the systems they compare.","excerpt":"The resulting retriever retrieves complete evidence more often than teacher-forced training and the systems we compare","url":"https://arxiv.org/abs/2610.40409","relationship":"Source-stated"},{"text":"The authors report transfer without further training to harder long-chain variants of HotpotQA and 2WikiMultiHopQA that they construct.","excerpt":"it transfers without further training to harder long-chain variants of HotpotQA and 2WikiMultiHopQA that we construct","url":"https://arxiv.org/abs/2610.40409","relationship":"Source-stated"},{"text":"The paper’s approach targets the problem that a retriever’s own earlier mistakes can make later missing evidence harder to retrieve.","excerpt":"these passages lower its chance of retrieving a missing gold passage well beyond what the difficulty of the states explains","url":"https://arxiv.org/abs/2610.40409","relationship":"Escape Meridian interpretation"}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2610.40409","checkedAt":"2026-11-13T09:10:07.394Z","items":[],"note":"No suitable media was found by the supported page extraction."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"How large are the reported gains, and how do they vary across the evaluated datasets and systems?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"Retrieving from Your Own Mistakes: On-Policy Training of Embedding Models for Multi-Hop Retrieval-Augmented Generation","summary":"In their paper, Chenjie Qiu and coauthors argue that a retriever for multi-hop retrieval-augmented generation should train on the retrieval states it reaches itself, including states containing mistakes. Chain embedding retrievers find evidence step by step, updating the question representation with passages already retrieved. The authors say training only on gold histories leaves a mismatch: at test time, a retriever conditions on its own passages, and by the third step about half or more of its states contain a wrong passage. Their on-policy state training approach labels each visited state with the earliest missing gold passage, masks other missing passages, and trains on these states alongside teacher-forced ones. The authors report that this retrieves complete evidence more often than teacher-forced training and compared systems, transfers without further training to harder long-chain variants of HotpotQA and 2WikiMultiHopQA, and recovers from its own mistakes more often on the same states. The idea matters because errors early in a multi-step search can make later evidence harder to find; the paper tests whether practicing on those error-shaped states helps the retriever recover. The abstract does not give the size of the gains or enough detail to assess their practical significance. How large are the reported gains, and how do they vary across the evaluated datasets and.","question":"How large are the reported gains, and how do they vary across the evaluated datasets and systems?","claims":[{"text":"The authors say a retriever trained only on gold histories can face a mismatch at test time, when it conditions on its own retrieved passages, which may be wrong.","excerpt":"At test time the retriever conditions on its own retrievals, which soon contain mistakes","support":"source-stated"},{"text":"The authors report that by the third step, about half or more of the states visited contain a wrong passage, and that these passages reduce the chance of retrieving missing gold evidence.","excerpt":"by the third step, about half or more of the states it visits hold a wrong passage, and these passages lower its chance of retrieving a missing gold passage","support":"source-stated"},{"text":"The proposed on-policy state training trains the encoder using states reached by its current retrievals, including states built from its own mistakes.","excerpt":"We propose on-policy state training, which trains the encoder on the states it reaches itself.","support":"source-stated"},{"text":"The authors report that the resulting retriever retrieves complete evidence more often than teacher-forced training and the systems they compare.","excerpt":"The resulting retriever retrieves complete evidence more often than teacher-forced training and the systems we compare","support":"source-stated"},{"text":"The authors report transfer without further training to harder long-chain variants of HotpotQA and 2WikiMultiHopQA that they construct.","excerpt":"it transfers without further training to harder long-chain variants of HotpotQA and 2WikiMultiHopQA that we construct","support":"source-stated"},{"text":"The paper’s approach targets the problem that a retriever’s own earlier mistakes can make later missing evidence harder to retrieve.","excerpt":"these passages lower its chance of retrieving a missing gold passage well beyond what the difficulty of the states explains","support":"interpretation"}],"limitations":["The supplied evidence is the paper’s abstract, which does not state the size of the reported gains or provide detailed results by dataset or comparison system.","The abstract describes harder long-chain variants of HotpotQA and 2WikiMultiHopQA that the authors construct; the supplied evidence does not describe those variants in detail."],"mediaId":"1c020a99b0a958ac24df25bff425fc078015c8ca6a9c4453f3e4cc6bc81ef9b2","correctionReason":null,"id":"b17157a7f4f43f7dde53453c3116a873f032f6d2edeac3ae07b9371b32f5733b","publicationId":"d08256f4-99a6-41f9-817f-b6ef0a597c9c","version":1,"sourceUrl":"https://arxiv.org/abs/2610.40409","sourceTitle":"Retrieving from Your Own Mistakes: On-Policy Training of Embedding Models for Multi-Hop Retrieval-Augmented Generation","publisher":"arXiv cs.AI","publishedAt":null,"originalPublished":{"value":"2026-10-29","precision":"day","basis":"publisher_citation_date"},"firstObservedAt":"2026-10-30T06:42:53.592Z","createdAt":"2026-11-13T07:03:07.331Z","historical":true,"entityIds":["arxiv"],"topicIds":["intelligence-and-discovery"],"media":[{"id":"1c020a99b0a958ac24df25bff425fc078015c8ca6a9c4453f3e4cc6bc81ef9b2","kind":"article","url":"https://arxiv.org/abs/2610.40409","sourceUrl":"https://arxiv.org/abs/2610.40409","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-10-31T09:31:07.096Z","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/2610.40409","checkedAt":"2026-11-13T09:10:07.394Z","items":[],"note":"No suitable media was found by the supported page extraction."},"editorialDecision":{"outcome":"accept","reasonCodes":["material_development"],"rubricVersion":"stream-admission-v5","mediaId":"1c020a99b0a958ac24df25bff425fc078015c8ca6a9c4453f3e4cc6bc81ef9b2","confidence":0.8200000000000001,"provider":"typesafe","model":"jev-1.13.0","inputHash":"4f8f5d529ca5b7183973fd94914ecdb2578baec2efb7b3f9908e106d1f97e61b","decidedAt":"2026-10-31T09:31:07.910Z"},"provenance":{"textHash":"c4b521a32436de807e84e36c0822b4ffd39efd14d51f2f3e33f7aef0faa335b4","observationId":"8bdfd9e8beeb32c2897f4e18ecd0e7e6a0af4b3557f428dc243ebfae7f8bb900","retrievedAt":"2026-10-31T09:31:07.007Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"428f50fe-2f2a-4708-8591-804a37db19ee"},"reviewer":{"model":"gpt-6-luna","runId":"f541e54d-6043-4703-8fb6-3d265597c7e9"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-13T07:03:07.331Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Qiu, Chenjie","url":null,"kind":"unknown"},{"name":"Zeng, Zixuan","url":null,"kind":"unknown"},{"name":"Shi, Shaobin","url":null,"kind":"unknown"},{"name":"Chen, Bingfeng","url":null,"kind":"unknown"},{"name":"Luo, Dongsheng","url":null,"kind":"unknown"},{"name":"Xu, Boyan","url":null,"kind":"unknown"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Qiu, Chenjie","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Qiu, Chenjie","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.40409","locator":"head/meta[name=author|citation_author]"},"assertionId":"b6a230068294134278e63603d3d8b19af6c1d0a47403ed347d46a1d1158a191c:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-13T09:10:07.394Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"b6a230068294134278e63603d3d8b19af6c1d0a47403ed347d46a1d1158a191c","sourceUrl":"https://arxiv.org/abs/2610.40409","contentSha256":"17a8ca064e4e27d1c3169945ee09ae8d5f7c7f64e10dc833acd6f31b6db24dc5","textHash":"c4b521a32436de807e84e36c0822b4ffd39efd14d51f2f3e33f7aef0faa335b4","observedAt":"2026-11-13T09:10:07.394Z"}},{"actor":{"name":"Zeng, Zixuan","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Zeng, Zixuan","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.40409","locator":"head/meta[name=author|citation_author]"},"assertionId":"b6a230068294134278e63603d3d8b19af6c1d0a47403ed347d46a1d1158a191c:participation:1","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-13T09:10:07.394Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"b6a230068294134278e63603d3d8b19af6c1d0a47403ed347d46a1d1158a191c","sourceUrl":"https://arxiv.org/abs/2610.40409","contentSha256":"17a8ca064e4e27d1c3169945ee09ae8d5f7c7f64e10dc833acd6f31b6db24dc5","textHash":"c4b521a32436de807e84e36c0822b4ffd39efd14d51f2f3e33f7aef0faa335b4","observedAt":"2026-11-13T09:10:07.394Z"}},{"actor":{"name":"Shi, Shaobin","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Shi, Shaobin","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.40409","locator":"head/meta[name=author|citation_author]"},"assertionId":"b6a230068294134278e63603d3d8b19af6c1d0a47403ed347d46a1d1158a191c:participation:2","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-13T09:10:07.394Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"b6a230068294134278e63603d3d8b19af6c1d0a47403ed347d46a1d1158a191c","sourceUrl":"https://arxiv.org/abs/2610.40409","contentSha256":"17a8ca064e4e27d1c3169945ee09ae8d5f7c7f64e10dc833acd6f31b6db24dc5","textHash":"c4b521a32436de807e84e36c0822b4ffd39efd14d51f2f3e33f7aef0faa335b4","observedAt":"2026-11-13T09:10:07.394Z"}},{"actor":{"name":"Chen, Bingfeng","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Chen, Bingfeng","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.40409","locator":"head/meta[name=author|citation_author]"},"assertionId":"b6a230068294134278e63603d3d8b19af6c1d0a47403ed347d46a1d1158a191c:participation:3","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-13T09:10:07.394Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"b6a230068294134278e63603d3d8b19af6c1d0a47403ed347d46a1d1158a191c","sourceUrl":"https://arxiv.org/abs/2610.40409","contentSha256":"17a8ca064e4e27d1c3169945ee09ae8d5f7c7f64e10dc833acd6f31b6db24dc5","textHash":"c4b521a32436de807e84e36c0822b4ffd39efd14d51f2f3e33f7aef0faa335b4","observedAt":"2026-11-13T09:10:07.394Z"}},{"actor":{"name":"Luo, Dongsheng","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Luo, Dongsheng","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.40409","locator":"head/meta[name=author|citation_author]"},"assertionId":"b6a230068294134278e63603d3d8b19af6c1d0a47403ed347d46a1d1158a191c:participation:4","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-13T09:10:07.394Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"b6a230068294134278e63603d3d8b19af6c1d0a47403ed347d46a1d1158a191c","sourceUrl":"https://arxiv.org/abs/2610.40409","contentSha256":"17a8ca064e4e27d1c3169945ee09ae8d5f7c7f64e10dc833acd6f31b6db24dc5","textHash":"c4b521a32436de807e84e36c0822b4ffd39efd14d51f2f3e33f7aef0faa335b4","observedAt":"2026-11-13T09:10:07.394Z"}},{"actor":{"name":"Xu, Boyan","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Xu, Boyan","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.40409","locator":"head/meta[name=author|citation_author]"},"assertionId":"b6a230068294134278e63603d3d8b19af6c1d0a47403ed347d46a1d1158a191c:participation:5","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-13T09:10:07.394Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"b6a230068294134278e63603d3d8b19af6c1d0a47403ed347d46a1d1158a191c","sourceUrl":"https://arxiv.org/abs/2610.40409","contentSha256":"17a8ca064e4e27d1c3169945ee09ae8d5f7c7f64e10dc833acd6f31b6db24dc5","textHash":"c4b521a32436de807e84e36c0822b4ffd39efd14d51f2f3e33f7aef0faa335b4","observedAt":"2026-11-13T09:10:07.394Z"}}],"participationCoverage":{"total":6,"returned":6,"complete":true,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":null,"sourceUrl":"https://arxiv.org/abs/2610.40409","snapshotId":"b6a230068294134278e63603d3d8b19af6c1d0a47403ed347d46a1d1158a191c","contentSha256":"17a8ca064e4e27d1c3169945ee09ae8d5f7c7f64e10dc833acd6f31b6db24dc5","textHash":"c4b521a32436de807e84e36c0822b4ffd39efd14d51f2f3e33f7aef0faa335b4","observedAt":"2026-11-13T09:10:07.394Z","basis":"publisher-page-metadata","sourceLinks":44,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/b17157a7f4f43f7dde53453c3116a873f032f6d2edeac3ae07b9371b32f5733b/source"}}}],"audits":[{"publicationId":"d08256f4-99a6-41f9-817f-b6ef0a597c9c","verdict":"uncertain","findings":[{"dimension":"attribution","status":"pass","explanation":"The draft correctly attributes the paper to Chenjie Qiu and coauthors, matching the listed authors. It frames findings as the authors' reports and claims rather than as established facts.","excerpt":"Authors:Chenjie Qiu, Zixuan Zeng, Shaobin Shi, Bingfeng Chen, Dongsheng Luo, Boyan Xu"},{"dimension":"summary","status":"uncertain","explanation":"The summary ends with a truncated, duplicated question fragment ('...datasets and.'). Remove it, since the question field already carries it. Central terms are left unexplained for a general reader: retrieval-augmented generation, embedding, gold histories, teacher-forced, and contrastive. Add plain-language glosses or term tooltips. The draft also omits the authors' finding that a small chain retriever trained on gold histories, calling no language model, already beats graph-based systems and is competitive with a search agent. It softens 'far more often' to 'more often'. Restore both to keep the source's force.","excerpt":"a small chain embedding retriever that calls no language model outperforms graph-based systems and is competitive with a search agent"},{"dimension":"date","status":"pass","explanation":"originalPublished 2026-10-29 matches the arXiv v1 submission date. The feed date of Oct 31 is later and correctly not used.","excerpt":"[Submitted on 29 Oct 2026]"},{"dimension":"support","status":"pass","explanation":"Each claim matches its excerpt. The interpretation claim about early mistakes hurting later retrieval is grounded in the source and labeled as interpretation. The limitations correctly note that the abstract gives no gain magnitudes. The question follows from that gap and is not a leading claim. No unsupported benefits or costs are asserted.","excerpt":"these passages lower its chance of retrieving a missing gold passage well beyond what the difficulty of the states explains"},{"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":"2026-11-13T07:03:07.331Z","model":"anthropic/claude-opus-5.5"}],"correctionUrl":"https://escapemeridian.com/corrections/"}