{"id":"6d7be7ca6401ca53b3eca0d4e9488fc8b93c8b07a76e751e5b1bc03dafc7b377","title":"arXiv cs.AI: WorkFactory turns professional intents into executable tasks for agent training","argument":"The authors of WorkFactory propose a framework for turning professional intents into tasks that can be executed to produce training examples. They say missing or inconsistent task information can disrupt teacher execution and weaken the resulting supervision. Their approach uses information-level planning to build an information support graph, which records information dependencies and locations needed for intended outputs. Probe executions identify failures in that support and guide repairs while preserving the intended work.\n\nThe authors report constructing 112,686 tasks across 612 occupations and using teacher trajectories to fine-tune Qwen models into the WorkPilot family. They report improvements over the corresponding base models on four professional-work benchmarks. On GDPval-AA v2.1, WorkPilot-35B-A3B rose from 879 to 1565 Elo. The authors also report that planning and repair reduce probe-diagnosed task environment defects and improve downstream performance.\n\nThese results suggest a way to scale the creation of professional-work training tasks, but the abstract does not establish how well the models perform in real workplaces. How do the reported benchmark results compare with performance on actual professional work?","topicIds":["intelligence-and-discovery"],"domain":"science","claimType":"attributed reading","limits":"The evidence is the paper's arXiv abstract; it does not provide benchmark methods, results by benchmark, or details needed to independently assess the reported findings. The abstract reports benchmark improvements but does not establish performance in real workplaces.","speaker":{"name":"Zeng, Ziyang, Chen, Tingyang, Deng, Jia, Wang, Qiyao, Tang, Shuo, Yang, Yike, Ren, Tao, Zhao, Wayne Xin, Chang, Feng, He, Jujie, Hao, Chuan, Tao, Ran, Dai, Bryan","roleAtSource":"Article author"},"source":{"title":"WorkFactory: Turning Professional Intents into Executable Work for Scalable Agent Training","publisher":"arXiv cs.AI","format":"Article","url":"https://arxiv.org/abs/2610.37161","publishedAt":"2026-10-27","dateBasis":"Date basis not provided."},"evidence":[{"text":"The authors propose WorkFactory to construct coherent, executable professional tasks from professional intents, using planning and probe-guided repair.","excerpt":"We propose WorkFactory, a scalable framework for constructing coherent and executable professional tasks.","url":"https://arxiv.org/abs/2610.37161","relationship":"Source-stated"},{"text":"The framework builds an information support graph to represent the information dependencies and locations needed for intended outputs.","excerpt":"explicitly representing the dependencies and locations of information required for the intended outputs.","url":"https://arxiv.org/abs/2610.37161","relationship":"Source-stated"},{"text":"The authors report building 112,686 tasks across 612 occupations and using teacher trajectories to fine-tune Qwen models into the WorkPilot family.","excerpt":"We construct 112,686 tasks spanning 612 occupations and use the collected teacher trajectories to fine-tune Qwen models","url":"https://arxiv.org/abs/2610.37161","relationship":"Source-stated"},{"text":"The authors report that all three WorkPilot models improve over their respective base models on four professional-work benchmarks.","excerpt":"All three models improve over their respective base models across four professional-work benchmarks","url":"https://arxiv.org/abs/2610.37161","relationship":"Source-stated"},{"text":"The authors report that WorkPilot-35B-A3B increased from 879 to 1565 Elo on GDPval-AA v2.1.","excerpt":"WorkPilot-35B-A3B improving from 879 to 1565 Elo on GDPval-AA v2.1.","url":"https://arxiv.org/abs/2610.37161","relationship":"Source-stated"},{"text":"The authors report that planning and repair reduce task environment defects identified by probes and improve downstream performance.","excerpt":"planning and repair reduce probe-diagnosed task environment defects and improve downstream performance.","url":"https://arxiv.org/abs/2610.37161","relationship":"Source-stated"}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2610.37161","checkedAt":"2026-11-14T06:48:13.878Z","items":[],"note":"No suitable media was found by the supported page extraction."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"How do the reported benchmark results compare with performance on actual professional work?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"arXiv cs.AI: WorkFactory turns professional intents into executable tasks for agent training","summary":"The authors of WorkFactory propose a framework for turning professional intents into tasks that can be executed to produce training examples. They say missing or inconsistent task information can disrupt teacher execution and weaken the resulting supervision. Their approach uses information-level planning to build an information support graph, which records information dependencies and locations needed for intended outputs. Probe executions identify failures in that support and guide repairs while preserving the intended work.\n\nThe authors report constructing 112,686 tasks across 612 occupations and using teacher trajectories to fine-tune Qwen models into the WorkPilot family. They report improvements over the corresponding base models on four professional-work benchmarks. On GDPval-AA v2.1, WorkPilot-35B-A3B rose from 879 to 1565 Elo. The authors also report that planning and repair reduce probe-diagnosed task environment defects and improve downstream performance.\n\nThese results suggest a way to scale the creation of professional-work training tasks, but the abstract does not establish how well the models perform in real workplaces. How do the reported benchmark results compare with performance on actual professional work?","question":"How do the reported benchmark results compare with performance on actual professional work?","claims":[{"text":"The authors propose WorkFactory to construct coherent, executable professional tasks from professional intents, using planning and probe-guided repair.","excerpt":"We propose WorkFactory, a scalable framework for constructing coherent and executable professional tasks.","support":"source-stated"},{"text":"The framework builds an information support graph to represent the information dependencies and locations needed for intended outputs.","excerpt":"explicitly representing the dependencies and locations of information required for the intended outputs.","support":"source-stated"},{"text":"The authors report building 112,686 tasks across 612 occupations and using teacher trajectories to fine-tune Qwen models into the WorkPilot family.","excerpt":"We construct 112,686 tasks spanning 612 occupations and use the collected teacher trajectories to fine-tune Qwen models","support":"source-stated"},{"text":"The authors report that all three WorkPilot models improve over their respective base models on four professional-work benchmarks.","excerpt":"All three models improve over their respective base models across four professional-work benchmarks","support":"source-stated"},{"text":"The authors report that WorkPilot-35B-A3B increased from 879 to 1565 Elo on GDPval-AA v2.1.","excerpt":"WorkPilot-35B-A3B improving from 879 to 1565 Elo on GDPval-AA v2.1.","support":"source-stated"},{"text":"The authors report that planning and repair reduce task environment defects identified by probes and improve downstream performance.","excerpt":"planning and repair reduce probe-diagnosed task environment defects and improve downstream performance.","support":"source-stated"}],"limitations":["The evidence is the paper's arXiv abstract; it does not provide benchmark methods, results by benchmark, or details needed to independently assess the reported findings.","The abstract reports benchmark improvements but does not establish performance in real workplaces."],"mediaId":"dd0575522a3404790b00bedaab28113ff1ddffa375894b6fa70894cda349759f","correctionReason":null,"id":"6d7be7ca6401ca53b3eca0d4e9488fc8b93c8b07a76e751e5b1bc03dafc7b377","publicationId":"72e373b1-a697-4fe7-8645-1eaf26c07c90","version":1,"sourceUrl":"https://arxiv.org/abs/2610.37161","sourceTitle":"WorkFactory: Turning Professional Intents into Executable Work for Scalable Agent Training","publisher":"arXiv cs.AI","publishedAt":null,"originalPublished":{"value":"2026-10-27","precision":"day","basis":"publisher_citation_date"},"firstObservedAt":"2026-10-28T08:44:36.583Z","createdAt":"2026-11-14T06:33:42.053Z","historical":true,"entityIds":["arxiv"],"topicIds":["intelligence-and-discovery"],"media":[{"id":"dd0575522a3404790b00bedaab28113ff1ddffa375894b6fa70894cda349759f","kind":"article","url":"https://arxiv.org/abs/2610.37161","sourceUrl":"https://arxiv.org/abs/2610.37161","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-10-30T01:46:16.278Z","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/2610.37161","checkedAt":"2026-11-14T06:48:13.878Z","items":[],"note":"No suitable media was found by the supported page extraction."},"editorialDecision":{"outcome":"accept","reasonCodes":["material_development"],"rubricVersion":"stream-admission-v5","mediaId":"dd0575522a3404790b00bedaab28113ff1ddffa375894b6fa70894cda349759f","confidence":0.94,"provider":"typesafe","model":"jev-1.13.0","inputHash":"2230ba7ea3a305b3ee4d9146e41b905be1aa6ef25167c1565c32e1ac100c8c01","decidedAt":"2026-10-30T01:46:18.671Z"},"provenance":{"textHash":"84547288ff8a02db2dcb922a1d65a8057c7de28878f007ea73a095b59cdcb7df","observationId":"f59ef4bf0c3cfd23381b41cc803622212b87b7432d68dfbdc3f680a00c662d9c","retrievedAt":"2026-10-30T01:46:15.985Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"f85bda99-8ee6-42b8-8f1a-4c57facccd62"},"reviewer":{"model":"gpt-6-luna","runId":"96238188-0733-4da4-818d-cc79707ccf0c"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-14T06:33:42.053Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Zeng, Ziyang","url":null,"kind":"unknown"},{"name":"Chen, Tingyang","url":null,"kind":"unknown"},{"name":"Deng, Jia","url":null,"kind":"unknown"},{"name":"Wang, Qiyao","url":null,"kind":"unknown"},{"name":"Tang, Shuo","url":null,"kind":"unknown"},{"name":"Yang, Yike","url":null,"kind":"unknown"},{"name":"Ren, Tao","url":null,"kind":"unknown"},{"name":"Zhao, Wayne Xin","url":null,"kind":"unknown"},{"name":"Chang, Feng","url":null,"kind":"unknown"},{"name":"He, Jujie","url":null,"kind":"unknown"},{"name":"Hao, Chuan","url":null,"kind":"unknown"},{"name":"Tao, Ran","url":null,"kind":"unknown"},{"name":"Dai, Bryan","url":null,"kind":"unknown"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Zeng, Ziyang","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Zeng, Ziyang","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.37161","locator":"head/meta[name=author|citation_author]"},"assertionId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T01:46:15.985Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789","sourceUrl":"https://arxiv.org/abs/2610.37161","contentSha256":"672ab9e450c279c9f52dc2c42fa09194329f194abb4c0f09ef2f997ecd645d70","textHash":"84547288ff8a02db2dcb922a1d65a8057c7de28878f007ea73a095b59cdcb7df","observedAt":"2026-10-30T01:46:15.985Z"}},{"actor":{"name":"Chen, Tingyang","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Chen, Tingyang","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.37161","locator":"head/meta[name=author|citation_author]"},"assertionId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789:participation:1","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T01:46:15.985Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789","sourceUrl":"https://arxiv.org/abs/2610.37161","contentSha256":"672ab9e450c279c9f52dc2c42fa09194329f194abb4c0f09ef2f997ecd645d70","textHash":"84547288ff8a02db2dcb922a1d65a8057c7de28878f007ea73a095b59cdcb7df","observedAt":"2026-10-30T01:46:15.985Z"}},{"actor":{"name":"Deng, Jia","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Deng, Jia","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.37161","locator":"head/meta[name=author|citation_author]"},"assertionId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789:participation:2","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T01:46:15.985Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789","sourceUrl":"https://arxiv.org/abs/2610.37161","contentSha256":"672ab9e450c279c9f52dc2c42fa09194329f194abb4c0f09ef2f997ecd645d70","textHash":"84547288ff8a02db2dcb922a1d65a8057c7de28878f007ea73a095b59cdcb7df","observedAt":"2026-10-30T01:46:15.985Z"}},{"actor":{"name":"Wang, Qiyao","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Wang, Qiyao","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.37161","locator":"head/meta[name=author|citation_author]"},"assertionId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789:participation:3","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T01:46:15.985Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789","sourceUrl":"https://arxiv.org/abs/2610.37161","contentSha256":"672ab9e450c279c9f52dc2c42fa09194329f194abb4c0f09ef2f997ecd645d70","textHash":"84547288ff8a02db2dcb922a1d65a8057c7de28878f007ea73a095b59cdcb7df","observedAt":"2026-10-30T01:46:15.985Z"}},{"actor":{"name":"Tang, Shuo","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Tang, Shuo","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.37161","locator":"head/meta[name=author|citation_author]"},"assertionId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789:participation:4","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T01:46:15.985Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789","sourceUrl":"https://arxiv.org/abs/2610.37161","contentSha256":"672ab9e450c279c9f52dc2c42fa09194329f194abb4c0f09ef2f997ecd645d70","textHash":"84547288ff8a02db2dcb922a1d65a8057c7de28878f007ea73a095b59cdcb7df","observedAt":"2026-10-30T01:46:15.985Z"}},{"actor":{"name":"Yang, Yike","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Yang, Yike","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.37161","locator":"head/meta[name=author|citation_author]"},"assertionId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789:participation:5","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T01:46:15.985Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789","sourceUrl":"https://arxiv.org/abs/2610.37161","contentSha256":"672ab9e450c279c9f52dc2c42fa09194329f194abb4c0f09ef2f997ecd645d70","textHash":"84547288ff8a02db2dcb922a1d65a8057c7de28878f007ea73a095b59cdcb7df","observedAt":"2026-10-30T01:46:15.985Z"}},{"actor":{"name":"Ren, Tao","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Ren, Tao","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.37161","locator":"head/meta[name=author|citation_author]"},"assertionId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789:participation:6","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T01:46:15.985Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789","sourceUrl":"https://arxiv.org/abs/2610.37161","contentSha256":"672ab9e450c279c9f52dc2c42fa09194329f194abb4c0f09ef2f997ecd645d70","textHash":"84547288ff8a02db2dcb922a1d65a8057c7de28878f007ea73a095b59cdcb7df","observedAt":"2026-10-30T01:46:15.985Z"}},{"actor":{"name":"Zhao, Wayne Xin","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Zhao, Wayne Xin","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.37161","locator":"head/meta[name=author|citation_author]"},"assertionId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789:participation:7","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T01:46:15.985Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789","sourceUrl":"https://arxiv.org/abs/2610.37161","contentSha256":"672ab9e450c279c9f52dc2c42fa09194329f194abb4c0f09ef2f997ecd645d70","textHash":"84547288ff8a02db2dcb922a1d65a8057c7de28878f007ea73a095b59cdcb7df","observedAt":"2026-10-30T01:46:15.985Z"}},{"actor":{"name":"Chang, Feng","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Chang, Feng","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.37161","locator":"head/meta[name=author|citation_author]"},"assertionId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789:participation:8","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T01:46:15.985Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789","sourceUrl":"https://arxiv.org/abs/2610.37161","contentSha256":"672ab9e450c279c9f52dc2c42fa09194329f194abb4c0f09ef2f997ecd645d70","textHash":"84547288ff8a02db2dcb922a1d65a8057c7de28878f007ea73a095b59cdcb7df","observedAt":"2026-10-30T01:46:15.985Z"}},{"actor":{"name":"He, Jujie","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"He, Jujie","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.37161","locator":"head/meta[name=author|citation_author]"},"assertionId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789:participation:9","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T01:46:15.985Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789","sourceUrl":"https://arxiv.org/abs/2610.37161","contentSha256":"672ab9e450c279c9f52dc2c42fa09194329f194abb4c0f09ef2f997ecd645d70","textHash":"84547288ff8a02db2dcb922a1d65a8057c7de28878f007ea73a095b59cdcb7df","observedAt":"2026-10-30T01:46:15.985Z"}},{"actor":{"name":"Hao, Chuan","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Hao, Chuan","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.37161","locator":"head/meta[name=author|citation_author]"},"assertionId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789:participation:10","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T01:46:15.985Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789","sourceUrl":"https://arxiv.org/abs/2610.37161","contentSha256":"672ab9e450c279c9f52dc2c42fa09194329f194abb4c0f09ef2f997ecd645d70","textHash":"84547288ff8a02db2dcb922a1d65a8057c7de28878f007ea73a095b59cdcb7df","observedAt":"2026-10-30T01:46:15.985Z"}},{"actor":{"name":"Tao, Ran","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Tao, Ran","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.37161","locator":"head/meta[name=author|citation_author]"},"assertionId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789:participation:11","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T01:46:15.985Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789","sourceUrl":"https://arxiv.org/abs/2610.37161","contentSha256":"672ab9e450c279c9f52dc2c42fa09194329f194abb4c0f09ef2f997ecd645d70","textHash":"84547288ff8a02db2dcb922a1d65a8057c7de28878f007ea73a095b59cdcb7df","observedAt":"2026-10-30T01:46:15.985Z"}},{"actor":{"name":"Dai, Bryan","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Dai, Bryan","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.37161","locator":"head/meta[name=author|citation_author]"},"assertionId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789:participation:12","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T01:46:15.985Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789","sourceUrl":"https://arxiv.org/abs/2610.37161","contentSha256":"672ab9e450c279c9f52dc2c42fa09194329f194abb4c0f09ef2f997ecd645d70","textHash":"84547288ff8a02db2dcb922a1d65a8057c7de28878f007ea73a095b59cdcb7df","observedAt":"2026-10-30T01:46:15.985Z"}}],"participationCoverage":{"total":13,"returned":13,"complete":true,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":null,"sourceUrl":"https://arxiv.org/abs/2610.37161","snapshotId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789","contentSha256":"672ab9e450c279c9f52dc2c42fa09194329f194abb4c0f09ef2f997ecd645d70","textHash":"84547288ff8a02db2dcb922a1d65a8057c7de28878f007ea73a095b59cdcb7df","observedAt":"2026-10-30T01:46:15.985Z","basis":"publisher-page-metadata","sourceLinks":52,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/6d7be7ca6401ca53b3eca0d4e9488fc8b93c8b07a76e751e5b1bc03dafc7b377/source"}},"canonicalUrl":"https://escapemeridian.com/readings/arxiv-cs-ai-workfactory-turns-professional-intents-into-6d7be7ca","formats":{"html":"/readings/arxiv-cs-ai-workfactory-turns-professional-intents-into-6d7be7ca","markdown":"/readings/arxiv-cs-ai-workfactory-turns-professional-intents-into-6d7be7ca.md","json":"/readings/arxiv-cs-ai-workfactory-turns-professional-intents-into-6d7be7ca.json"},"attribution":"Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.","maintainedGuideLinks":[],"conditions":[],"readerContext":{"version":"reader-context-v1","attribution":"Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.","terms":[{"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"},{"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"}],"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":"6d7be7ca6401ca53b3eca0d4e9488fc8b93c8b07a76e751e5b1bc03dafc7b377","revision":"72e373b1-a697-4fe7-8645-1eaf26c07c90","revisionKind":"publication-id","title":"arXiv cs.AI: WorkFactory turns professional intents into executable tasks for agent training","url":"https://escapemeridian.com/readings/arxiv-cs-ai-workfactory-turns-professional-intents-into-6d7be7ca"},"source":{"type":"source","id":"https://arxiv.org/abs/2610.37161","revision":"84547288ff8a02db2dcb922a1d65a8057c7de28878f007ea73a095b59cdcb7df","revisionKind":"sha256-source-text","title":"WorkFactory: Turning Professional Intents into Executable Work for Scalable Agent Training","url":"https://arxiv.org/abs/2610.37161"},"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":"6d7be7ca6401ca53b3eca0d4e9488fc8b93c8b07a76e751e5b1bc03dafc7b377","title":"arXiv cs.AI: WorkFactory turns professional intents into executable tasks for agent training","argument":"The authors of WorkFactory propose a framework for turning professional intents into tasks that can be executed to produce training examples. They say missing or inconsistent task information can disrupt teacher execution and weaken the resulting supervision. Their approach uses information-level planning to build an information support graph, which records information dependencies and locations needed for intended outputs. Probe executions identify failures in that support and guide repairs while preserving the intended work.\n\nThe authors report constructing 112,686 tasks across 612 occupations and using teacher trajectories to fine-tune Qwen models into the WorkPilot family. They report improvements over the corresponding base models on four professional-work benchmarks. On GDPval-AA v2.1, WorkPilot-35B-A3B rose from 879 to 1565 Elo. The authors also report that planning and repair reduce probe-diagnosed task environment defects and improve downstream performance.\n\nThese results suggest a way to scale the creation of professional-work training tasks, but the abstract does not establish how well the models perform in real workplaces. How do the reported benchmark results compare with performance on actual professional work?","topicIds":["intelligence-and-discovery"],"domain":"science","claimType":"attributed reading","limits":"The evidence is the paper's arXiv abstract; it does not provide benchmark methods, results by benchmark, or details needed to independently assess the reported findings. The abstract reports benchmark improvements but does not establish performance in real workplaces.","speaker":{"name":"Zeng, Ziyang, Chen, Tingyang, Deng, Jia, Wang, Qiyao, Tang, Shuo, Yang, Yike, Ren, Tao, Zhao, Wayne Xin, Chang, Feng, He, Jujie, Hao, Chuan, Tao, Ran, Dai, Bryan","roleAtSource":"Article author"},"source":{"title":"WorkFactory: Turning Professional Intents into Executable Work for Scalable Agent Training","publisher":"arXiv cs.AI","format":"Article","url":"https://arxiv.org/abs/2610.37161","publishedAt":"2026-10-27","dateBasis":"Date basis not provided."},"evidence":[{"text":"The authors propose WorkFactory to construct coherent, executable professional tasks from professional intents, using planning and probe-guided repair.","excerpt":"We propose WorkFactory, a scalable framework for constructing coherent and executable professional tasks.","url":"https://arxiv.org/abs/2610.37161","relationship":"Source-stated"},{"text":"The framework builds an information support graph to represent the information dependencies and locations needed for intended outputs.","excerpt":"explicitly representing the dependencies and locations of information required for the intended outputs.","url":"https://arxiv.org/abs/2610.37161","relationship":"Source-stated"},{"text":"The authors report building 112,686 tasks across 612 occupations and using teacher trajectories to fine-tune Qwen models into the WorkPilot family.","excerpt":"We construct 112,686 tasks spanning 612 occupations and use the collected teacher trajectories to fine-tune Qwen models","url":"https://arxiv.org/abs/2610.37161","relationship":"Source-stated"},{"text":"The authors report that all three WorkPilot models improve over their respective base models on four professional-work benchmarks.","excerpt":"All three models improve over their respective base models across four professional-work benchmarks","url":"https://arxiv.org/abs/2610.37161","relationship":"Source-stated"},{"text":"The authors report that WorkPilot-35B-A3B increased from 879 to 1565 Elo on GDPval-AA v2.1.","excerpt":"WorkPilot-35B-A3B improving from 879 to 1565 Elo on GDPval-AA v2.1.","url":"https://arxiv.org/abs/2610.37161","relationship":"Source-stated"},{"text":"The authors report that planning and repair reduce task environment defects identified by probes and improve downstream performance.","excerpt":"planning and repair reduce probe-diagnosed task environment defects and improve downstream performance.","url":"https://arxiv.org/abs/2610.37161","relationship":"Source-stated"}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2610.37161","checkedAt":"2026-11-14T06:48:13.878Z","items":[],"note":"No suitable media was found by the supported page extraction."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"How do the reported benchmark results compare with performance on actual professional work?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"arXiv cs.AI: WorkFactory turns professional intents into executable tasks for agent training","summary":"The authors of WorkFactory propose a framework for turning professional intents into tasks that can be executed to produce training examples. They say missing or inconsistent task information can disrupt teacher execution and weaken the resulting supervision. Their approach uses information-level planning to build an information support graph, which records information dependencies and locations needed for intended outputs. Probe executions identify failures in that support and guide repairs while preserving the intended work.\n\nThe authors report constructing 112,686 tasks across 612 occupations and using teacher trajectories to fine-tune Qwen models into the WorkPilot family. They report improvements over the corresponding base models on four professional-work benchmarks. On GDPval-AA v2.1, WorkPilot-35B-A3B rose from 879 to 1565 Elo. The authors also report that planning and repair reduce probe-diagnosed task environment defects and improve downstream performance.\n\nThese results suggest a way to scale the creation of professional-work training tasks, but the abstract does not establish how well the models perform in real workplaces. How do the reported benchmark results compare with performance on actual professional work?","question":"How do the reported benchmark results compare with performance on actual professional work?","claims":[{"text":"The authors propose WorkFactory to construct coherent, executable professional tasks from professional intents, using planning and probe-guided repair.","excerpt":"We propose WorkFactory, a scalable framework for constructing coherent and executable professional tasks.","support":"source-stated"},{"text":"The framework builds an information support graph to represent the information dependencies and locations needed for intended outputs.","excerpt":"explicitly representing the dependencies and locations of information required for the intended outputs.","support":"source-stated"},{"text":"The authors report building 112,686 tasks across 612 occupations and using teacher trajectories to fine-tune Qwen models into the WorkPilot family.","excerpt":"We construct 112,686 tasks spanning 612 occupations and use the collected teacher trajectories to fine-tune Qwen models","support":"source-stated"},{"text":"The authors report that all three WorkPilot models improve over their respective base models on four professional-work benchmarks.","excerpt":"All three models improve over their respective base models across four professional-work benchmarks","support":"source-stated"},{"text":"The authors report that WorkPilot-35B-A3B increased from 879 to 1565 Elo on GDPval-AA v2.1.","excerpt":"WorkPilot-35B-A3B improving from 879 to 1565 Elo on GDPval-AA v2.1.","support":"source-stated"},{"text":"The authors report that planning and repair reduce task environment defects identified by probes and improve downstream performance.","excerpt":"planning and repair reduce probe-diagnosed task environment defects and improve downstream performance.","support":"source-stated"}],"limitations":["The evidence is the paper's arXiv abstract; it does not provide benchmark methods, results by benchmark, or details needed to independently assess the reported findings.","The abstract reports benchmark improvements but does not establish performance in real workplaces."],"mediaId":"dd0575522a3404790b00bedaab28113ff1ddffa375894b6fa70894cda349759f","correctionReason":null,"id":"6d7be7ca6401ca53b3eca0d4e9488fc8b93c8b07a76e751e5b1bc03dafc7b377","publicationId":"72e373b1-a697-4fe7-8645-1eaf26c07c90","version":1,"sourceUrl":"https://arxiv.org/abs/2610.37161","sourceTitle":"WorkFactory: Turning Professional Intents into Executable Work for Scalable Agent Training","publisher":"arXiv cs.AI","publishedAt":null,"originalPublished":{"value":"2026-10-27","precision":"day","basis":"publisher_citation_date"},"firstObservedAt":"2026-10-28T08:44:36.583Z","createdAt":"2026-11-14T06:33:42.053Z","historical":true,"entityIds":["arxiv"],"topicIds":["intelligence-and-discovery"],"media":[{"id":"dd0575522a3404790b00bedaab28113ff1ddffa375894b6fa70894cda349759f","kind":"article","url":"https://arxiv.org/abs/2610.37161","sourceUrl":"https://arxiv.org/abs/2610.37161","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-10-30T01:46:16.278Z","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/2610.37161","checkedAt":"2026-11-14T06:48:13.878Z","items":[],"note":"No suitable media was found by the supported page extraction."},"editorialDecision":{"outcome":"accept","reasonCodes":["material_development"],"rubricVersion":"stream-admission-v5","mediaId":"dd0575522a3404790b00bedaab28113ff1ddffa375894b6fa70894cda349759f","confidence":0.94,"provider":"typesafe","model":"jev-1.13.0","inputHash":"2230ba7ea3a305b3ee4d9146e41b905be1aa6ef25167c1565c32e1ac100c8c01","decidedAt":"2026-10-30T01:46:18.671Z"},"provenance":{"textHash":"84547288ff8a02db2dcb922a1d65a8057c7de28878f007ea73a095b59cdcb7df","observationId":"f59ef4bf0c3cfd23381b41cc803622212b87b7432d68dfbdc3f680a00c662d9c","retrievedAt":"2026-10-30T01:46:15.985Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"f85bda99-8ee6-42b8-8f1a-4c57facccd62"},"reviewer":{"model":"gpt-6-luna","runId":"96238188-0733-4da4-818d-cc79707ccf0c"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-14T06:33:42.053Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Zeng, Ziyang","url":null,"kind":"unknown"},{"name":"Chen, Tingyang","url":null,"kind":"unknown"},{"name":"Deng, Jia","url":null,"kind":"unknown"},{"name":"Wang, Qiyao","url":null,"kind":"unknown"},{"name":"Tang, Shuo","url":null,"kind":"unknown"},{"name":"Yang, Yike","url":null,"kind":"unknown"},{"name":"Ren, Tao","url":null,"kind":"unknown"},{"name":"Zhao, Wayne Xin","url":null,"kind":"unknown"},{"name":"Chang, Feng","url":null,"kind":"unknown"},{"name":"He, Jujie","url":null,"kind":"unknown"},{"name":"Hao, Chuan","url":null,"kind":"unknown"},{"name":"Tao, Ran","url":null,"kind":"unknown"},{"name":"Dai, Bryan","url":null,"kind":"unknown"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Zeng, Ziyang","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Zeng, Ziyang","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.37161","locator":"head/meta[name=author|citation_author]"},"assertionId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T01:46:15.985Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789","sourceUrl":"https://arxiv.org/abs/2610.37161","contentSha256":"672ab9e450c279c9f52dc2c42fa09194329f194abb4c0f09ef2f997ecd645d70","textHash":"84547288ff8a02db2dcb922a1d65a8057c7de28878f007ea73a095b59cdcb7df","observedAt":"2026-10-30T01:46:15.985Z"}},{"actor":{"name":"Chen, Tingyang","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Chen, Tingyang","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.37161","locator":"head/meta[name=author|citation_author]"},"assertionId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789:participation:1","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T01:46:15.985Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789","sourceUrl":"https://arxiv.org/abs/2610.37161","contentSha256":"672ab9e450c279c9f52dc2c42fa09194329f194abb4c0f09ef2f997ecd645d70","textHash":"84547288ff8a02db2dcb922a1d65a8057c7de28878f007ea73a095b59cdcb7df","observedAt":"2026-10-30T01:46:15.985Z"}},{"actor":{"name":"Deng, Jia","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Deng, Jia","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.37161","locator":"head/meta[name=author|citation_author]"},"assertionId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789:participation:2","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T01:46:15.985Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789","sourceUrl":"https://arxiv.org/abs/2610.37161","contentSha256":"672ab9e450c279c9f52dc2c42fa09194329f194abb4c0f09ef2f997ecd645d70","textHash":"84547288ff8a02db2dcb922a1d65a8057c7de28878f007ea73a095b59cdcb7df","observedAt":"2026-10-30T01:46:15.985Z"}},{"actor":{"name":"Wang, Qiyao","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Wang, Qiyao","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.37161","locator":"head/meta[name=author|citation_author]"},"assertionId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789:participation:3","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T01:46:15.985Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789","sourceUrl":"https://arxiv.org/abs/2610.37161","contentSha256":"672ab9e450c279c9f52dc2c42fa09194329f194abb4c0f09ef2f997ecd645d70","textHash":"84547288ff8a02db2dcb922a1d65a8057c7de28878f007ea73a095b59cdcb7df","observedAt":"2026-10-30T01:46:15.985Z"}},{"actor":{"name":"Tang, Shuo","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Tang, Shuo","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.37161","locator":"head/meta[name=author|citation_author]"},"assertionId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789:participation:4","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T01:46:15.985Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789","sourceUrl":"https://arxiv.org/abs/2610.37161","contentSha256":"672ab9e450c279c9f52dc2c42fa09194329f194abb4c0f09ef2f997ecd645d70","textHash":"84547288ff8a02db2dcb922a1d65a8057c7de28878f007ea73a095b59cdcb7df","observedAt":"2026-10-30T01:46:15.985Z"}},{"actor":{"name":"Yang, Yike","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Yang, Yike","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.37161","locator":"head/meta[name=author|citation_author]"},"assertionId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789:participation:5","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T01:46:15.985Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789","sourceUrl":"https://arxiv.org/abs/2610.37161","contentSha256":"672ab9e450c279c9f52dc2c42fa09194329f194abb4c0f09ef2f997ecd645d70","textHash":"84547288ff8a02db2dcb922a1d65a8057c7de28878f007ea73a095b59cdcb7df","observedAt":"2026-10-30T01:46:15.985Z"}},{"actor":{"name":"Ren, Tao","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Ren, Tao","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.37161","locator":"head/meta[name=author|citation_author]"},"assertionId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789:participation:6","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T01:46:15.985Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789","sourceUrl":"https://arxiv.org/abs/2610.37161","contentSha256":"672ab9e450c279c9f52dc2c42fa09194329f194abb4c0f09ef2f997ecd645d70","textHash":"84547288ff8a02db2dcb922a1d65a8057c7de28878f007ea73a095b59cdcb7df","observedAt":"2026-10-30T01:46:15.985Z"}},{"actor":{"name":"Zhao, Wayne Xin","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Zhao, Wayne Xin","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.37161","locator":"head/meta[name=author|citation_author]"},"assertionId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789:participation:7","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T01:46:15.985Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789","sourceUrl":"https://arxiv.org/abs/2610.37161","contentSha256":"672ab9e450c279c9f52dc2c42fa09194329f194abb4c0f09ef2f997ecd645d70","textHash":"84547288ff8a02db2dcb922a1d65a8057c7de28878f007ea73a095b59cdcb7df","observedAt":"2026-10-30T01:46:15.985Z"}},{"actor":{"name":"Chang, Feng","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Chang, Feng","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.37161","locator":"head/meta[name=author|citation_author]"},"assertionId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789:participation:8","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T01:46:15.985Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789","sourceUrl":"https://arxiv.org/abs/2610.37161","contentSha256":"672ab9e450c279c9f52dc2c42fa09194329f194abb4c0f09ef2f997ecd645d70","textHash":"84547288ff8a02db2dcb922a1d65a8057c7de28878f007ea73a095b59cdcb7df","observedAt":"2026-10-30T01:46:15.985Z"}},{"actor":{"name":"He, Jujie","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"He, Jujie","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.37161","locator":"head/meta[name=author|citation_author]"},"assertionId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789:participation:9","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T01:46:15.985Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789","sourceUrl":"https://arxiv.org/abs/2610.37161","contentSha256":"672ab9e450c279c9f52dc2c42fa09194329f194abb4c0f09ef2f997ecd645d70","textHash":"84547288ff8a02db2dcb922a1d65a8057c7de28878f007ea73a095b59cdcb7df","observedAt":"2026-10-30T01:46:15.985Z"}},{"actor":{"name":"Hao, Chuan","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Hao, Chuan","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.37161","locator":"head/meta[name=author|citation_author]"},"assertionId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789:participation:10","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T01:46:15.985Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789","sourceUrl":"https://arxiv.org/abs/2610.37161","contentSha256":"672ab9e450c279c9f52dc2c42fa09194329f194abb4c0f09ef2f997ecd645d70","textHash":"84547288ff8a02db2dcb922a1d65a8057c7de28878f007ea73a095b59cdcb7df","observedAt":"2026-10-30T01:46:15.985Z"}},{"actor":{"name":"Tao, Ran","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Tao, Ran","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.37161","locator":"head/meta[name=author|citation_author]"},"assertionId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789:participation:11","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T01:46:15.985Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789","sourceUrl":"https://arxiv.org/abs/2610.37161","contentSha256":"672ab9e450c279c9f52dc2c42fa09194329f194abb4c0f09ef2f997ecd645d70","textHash":"84547288ff8a02db2dcb922a1d65a8057c7de28878f007ea73a095b59cdcb7df","observedAt":"2026-10-30T01:46:15.985Z"}},{"actor":{"name":"Dai, Bryan","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Dai, Bryan","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.37161","locator":"head/meta[name=author|citation_author]"},"assertionId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789:participation:12","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T01:46:15.985Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789","sourceUrl":"https://arxiv.org/abs/2610.37161","contentSha256":"672ab9e450c279c9f52dc2c42fa09194329f194abb4c0f09ef2f997ecd645d70","textHash":"84547288ff8a02db2dcb922a1d65a8057c7de28878f007ea73a095b59cdcb7df","observedAt":"2026-10-30T01:46:15.985Z"}}],"participationCoverage":{"total":13,"returned":13,"complete":true,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":null,"sourceUrl":"https://arxiv.org/abs/2610.37161","snapshotId":"101f8f746bacea026e0238d40cf3dee3841ad86c2951b35c01f436b4db0e6789","contentSha256":"672ab9e450c279c9f52dc2c42fa09194329f194abb4c0f09ef2f997ecd645d70","textHash":"84547288ff8a02db2dcb922a1d65a8057c7de28878f007ea73a095b59cdcb7df","observedAt":"2026-10-30T01:46:15.985Z","basis":"publisher-page-metadata","sourceLinks":52,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/6d7be7ca6401ca53b3eca0d4e9488fc8b93c8b07a76e751e5b1bc03dafc7b377/source"}}}],"audits":[{"publicationId":"72e373b1-a697-4fe7-8645-1eaf26c07c90","verdict":"uncertain","findings":[{"dimension":"attribution","status":"pass","explanation":"The reading attributes the framework, task counts, model results and component analyses to the paper's authors. It keeps reported results as author claims and does not present them as Escape Meridian findings.","excerpt":"We propose WorkFactory, a scalable framework for constructing coherent and executable professional tasks."},{"dimension":"summary","status":"uncertain","explanation":"The summary is faithful but hard to follow without specialist background. It never says that a 'teacher' (presumably a stronger model) carries out the tasks and that its recorded trajectories become training data, and it does not explain what an Elo jump from 879 to 1565 means. It also drops the authors' scaling result: diminishing returns that vary by model, with execution length first rising then falling. Smallest repair: add one plain sentence on the teacher-demonstration mechanism and Elo, plus one sentence on the diminishing-returns finding, using term-library tooltips for central terms.","excerpt":"Scaling analyses reveal model-dependent diminishing returns, with execution length first increasing and then decreasing."},{"dimension":"date","status":"pass","explanation":"The originalPublished date of 2026-10-27 matches the v1 submission date shown on the arXiv page. It is not the later feed date of 2026-10-29.","excerpt":"[Submitted on 27 Oct 2026]"},{"dimension":"support","status":"pass","explanation":"Each claim matches the abstract verbatim in substance. The closing line about suggesting a way to scale task creation follows the authors' own 'scalable' framing, and it is properly hedged: the reading notes that real-workplace performance is not established. The question follows from that evidence gap and does not lead the reader.","excerpt":"All three models improve over their respective base models across four professional-work benchmarks, with WorkPilot-35B-A3B improving from 879 to 1565 Elo on GDPval-AA v2.1."},{"dimension":"media","status":"pass","explanation":"The selected media is the verified arXiv abstract page for this paper. It is the direct original source and the easiest way for readers to reach it.","excerpt":null}],"createdAt":"2026-11-14T06:33:42.053Z","model":"anthropic/claude-opus-5.5"}],"correctionUrl":"https://escapemeridian.com/corrections/"}