{"id":"9fef612dfdc3e182b35132ab11b3a5cdeef567d527a6722399c44663e1cdbec5","title":"arXiv cs.AI authors propose MAREval to evaluate recommendation explanations","argument":"The authors of MAREval propose a framework that uses large language models to evaluate explanations produced by recommender systems. Their approach assigns different roles to agents: a planner coordinates the evaluation, a moderator helps limit prompt drift, and an arbitrator combines results across evaluation rounds. A Monte Carlo sampling method is also used to improve robustness and alignment with human judgment.\n\nThe authors report that MAREval outperformed existing methods on public and proprietary recommendation datasets, showed lower variability across repeated trials, and met production quality thresholds in a large human-annotation gate. They also report statistically significant gains in engagement and revenue metrics in an online A/B test. These are results reported in the paper’s abstract; the supplied evidence does not include study details needed to assess how broadly they apply.\n\nEscape Meridian interpretation: If the reported results generalize, MAREval could help recommendation platforms evaluate explanations at scale, where the authors say human evaluation is resource-intensive. How well does the framework perform across different platforms, user groups, and recommendation tasks?","topicIds":["intelligence-and-discovery"],"domain":"science","claimType":"attributed reading","limits":"The supplied evidence is the paper’s abstract, not its full methods or results; details needed to assess the reported evaluations are not included. The abstract reports results on public and proprietary datasets, but the supplied evidence does not establish how well they generalize across platforms, user groups, or recommendation tasks.","speaker":{"name":"Maragheh, Reza Yousefi, Kudase, Jayesh Uddhav, Inan, Aysenur, Giahi, Ramin, Zhao, Kai, Parikh, Parth, Xu, Jianpeng, Cho, Jason, Korpeoglu, Evren, Kumar, Sushant","roleAtSource":"Article author"},"source":{"title":"MAREval: A Multi-Agent Framework for Evaluating Natural Language Recommendation Explanations","publisher":"arXiv cs.AI","format":"Article","url":"https://arxiv.org/abs/2610.36389","publishedAt":"2026-10-26","dateBasis":"Date basis not provided."},"evidence":[{"text":"The authors propose MAREval, a multi-agent framework using large language models to evaluate explanations in recommender systems.","excerpt":"we propose MAREval, a structured multi-agent framework for evaluating recommendation explanations using large language models.","url":"https://arxiv.org/abs/2610.36389","relationship":"Source-stated"},{"text":"The framework uses planner, moderator, and arbitrator agents, along with Monte Carlo sampling.","excerpt":"a planner agent that uses a novel Chain of Debate (CoD) prompting strategy to coordinate agent roles","url":"https://arxiv.org/abs/2610.36389","relationship":"Source-stated"},{"text":"The authors report that MAREval outperformed state-of-the-art baselines on public and proprietary recommendation datasets.","excerpt":"demonstrating that MAREval outperforms state-of-the-art baselines.","url":"https://arxiv.org/abs/2610.36389","relationship":"Source-stated"},{"text":"The authors report lower variability across repeated trials and that MAREval met production quality thresholds in a large human-annotation gate.","excerpt":"Stability analyses indicate substantially lower variability across repeated trials. In a large human-annotation gate, MAREval meets production quality thresholds","url":"https://arxiv.org/abs/2610.36389","relationship":"Source-stated"},{"text":"The authors report statistically significant improvements in engagement and revenue metrics in online A/B testing.","excerpt":"online A/B testing demonstrates statistically significant improvements in engagement and revenue metrics.","url":"https://arxiv.org/abs/2610.36389","relationship":"Source-stated"},{"text":"If the reported results generalize, MAREval could help platforms evaluate recommendation explanations at scale.","excerpt":"human evaluations, while accurate, are resource-intensive and impractical at the scale required by modern recommendation platforms.","url":"https://arxiv.org/abs/2610.36389","relationship":"Escape Meridian interpretation"}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2610.36389","checkedAt":"2026-11-14T03:48:10.071Z","items":[],"note":"No suitable media was found by the supported page extraction."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"How well does MAREval perform across different platforms, user groups, and recommendation tasks?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"arXiv cs.AI authors propose MAREval to evaluate recommendation explanations","summary":"The authors of MAREval propose a framework that uses large language models to evaluate explanations produced by recommender systems. Their approach assigns different roles to agents: a planner coordinates the evaluation, a moderator helps limit prompt drift, and an arbitrator combines results across evaluation rounds. A Monte Carlo sampling method is also used to improve robustness and alignment with human judgment.\n\nThe authors report that MAREval outperformed existing methods on public and proprietary recommendation datasets, showed lower variability across repeated trials, and met production quality thresholds in a large human-annotation gate. They also report statistically significant gains in engagement and revenue metrics in an online A/B test. These are results reported in the paper’s abstract; the supplied evidence does not include study details needed to assess how broadly they apply.\n\nEscape Meridian interpretation: If the reported results generalize, MAREval could help recommendation platforms evaluate explanations at scale, where the authors say human evaluation is resource-intensive. How well does the framework perform across different platforms, user groups, and recommendation tasks?","question":"How well does MAREval perform across different platforms, user groups, and recommendation tasks?","claims":[{"text":"The authors propose MAREval, a multi-agent framework using large language models to evaluate explanations in recommender systems.","excerpt":"we propose MAREval, a structured multi-agent framework for evaluating recommendation explanations using large language models.","support":"source-stated"},{"text":"The framework uses planner, moderator, and arbitrator agents, along with Monte Carlo sampling.","excerpt":"a planner agent that uses a novel Chain of Debate (CoD) prompting strategy to coordinate agent roles","support":"source-stated"},{"text":"The authors report that MAREval outperformed state-of-the-art baselines on public and proprietary recommendation datasets.","excerpt":"demonstrating that MAREval outperforms state-of-the-art baselines.","support":"source-stated"},{"text":"The authors report lower variability across repeated trials and that MAREval met production quality thresholds in a large human-annotation gate.","excerpt":"Stability analyses indicate substantially lower variability across repeated trials. In a large human-annotation gate, MAREval meets production quality thresholds","support":"source-stated"},{"text":"The authors report statistically significant improvements in engagement and revenue metrics in online A/B testing.","excerpt":"online A/B testing demonstrates statistically significant improvements in engagement and revenue metrics.","support":"source-stated"},{"text":"If the reported results generalize, MAREval could help platforms evaluate recommendation explanations at scale.","excerpt":"human evaluations, while accurate, are resource-intensive and impractical at the scale required by modern recommendation platforms.","support":"interpretation"}],"limitations":["The supplied evidence is the paper’s abstract, not its full methods or results; details needed to assess the reported evaluations are not included.","The abstract reports results on public and proprietary datasets, but the supplied evidence does not establish how well they generalize across platforms, user groups, or recommendation tasks."],"mediaId":"123381003ba50a8e3fb7fd1a11c14fa8be2ab4fe4ad320983dd0eaa99ffd0e32","correctionReason":null,"id":"9fef612dfdc3e182b35132ab11b3a5cdeef567d527a6722399c44663e1cdbec5","publicationId":"5715dc27-2c68-49ee-86bf-4075d26bc49e","version":1,"sourceUrl":"https://arxiv.org/abs/2610.36389","sourceTitle":"MAREval: A Multi-Agent Framework for Evaluating Natural Language Recommendation Explanations","publisher":"arXiv cs.AI","publishedAt":null,"originalPublished":{"value":"2026-10-26","precision":"day","basis":"publisher_citation_date"},"firstObservedAt":"2026-10-28T08:04:31.779Z","createdAt":"2026-11-14T03:33:03.238Z","historical":true,"entityIds":["arxiv"],"topicIds":["intelligence-and-discovery"],"media":[{"id":"123381003ba50a8e3fb7fd1a11c14fa8be2ab4fe4ad320983dd0eaa99ffd0e32","kind":"article","url":"https://arxiv.org/abs/2610.36389","sourceUrl":"https://arxiv.org/abs/2610.36389","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-10-30T05:46:44.220Z","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/2610.36389","checkedAt":"2026-11-14T03:48:10.071Z","items":[],"note":"No suitable media was found by the supported page extraction."},"editorialDecision":{"outcome":"accept","reasonCodes":["material_development"],"rubricVersion":"stream-admission-v5","mediaId":"123381003ba50a8e3fb7fd1a11c14fa8be2ab4fe4ad320983dd0eaa99ffd0e32","confidence":0.8,"provider":"typesafe","model":"jev-1.13.0","inputHash":"8f0bb481efc577f1ae445cf8565642d18e01ec802627cd1fa3d34b911137fc25","decidedAt":"2026-10-30T05:46:44.896Z"},"provenance":{"textHash":"7446b11e2eea5aa06b84f006ee142e1899d35be991bfb92ffee32b329d720c2e","observationId":"768bc19feb8da9e4833470b388aa2116710d9fdf6acf848a1b56b734cc91566e","retrievedAt":"2026-10-30T05:46:44.117Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"ab8c9f98-1904-4af0-88a9-118298ca19bd"},"reviewer":{"model":"gpt-6-luna","runId":"06de8221-632c-4c86-873f-521dade2fe51"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-14T03:33:03.238Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Maragheh, Reza Yousefi","url":null,"kind":"unknown"},{"name":"Kudase, Jayesh Uddhav","url":null,"kind":"unknown"},{"name":"Inan, Aysenur","url":null,"kind":"unknown"},{"name":"Giahi, Ramin","url":null,"kind":"unknown"},{"name":"Zhao, Kai","url":null,"kind":"unknown"},{"name":"Parikh, Parth","url":null,"kind":"unknown"},{"name":"Xu, Jianpeng","url":null,"kind":"unknown"},{"name":"Cho, Jason","url":null,"kind":"unknown"},{"name":"Korpeoglu, Evren","url":null,"kind":"unknown"},{"name":"Kumar, Sushant","url":null,"kind":"unknown"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Maragheh, Reza Yousefi","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Maragheh, Reza Yousefi","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.36389","locator":"head/meta[name=author|citation_author]"},"assertionId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T05:46:44.117Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459","sourceUrl":"https://arxiv.org/abs/2610.36389","contentSha256":"7d188d1a852eb2dba8c04b9ea982f7f4870dd3547c392712f87065134ccfee98","textHash":"7446b11e2eea5aa06b84f006ee142e1899d35be991bfb92ffee32b329d720c2e","observedAt":"2026-10-30T05:46:44.117Z"}},{"actor":{"name":"Kudase, Jayesh Uddhav","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Kudase, Jayesh Uddhav","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.36389","locator":"head/meta[name=author|citation_author]"},"assertionId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459:participation:1","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T05:46:44.117Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459","sourceUrl":"https://arxiv.org/abs/2610.36389","contentSha256":"7d188d1a852eb2dba8c04b9ea982f7f4870dd3547c392712f87065134ccfee98","textHash":"7446b11e2eea5aa06b84f006ee142e1899d35be991bfb92ffee32b329d720c2e","observedAt":"2026-10-30T05:46:44.117Z"}},{"actor":{"name":"Inan, Aysenur","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Inan, Aysenur","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.36389","locator":"head/meta[name=author|citation_author]"},"assertionId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459:participation:2","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T05:46:44.117Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459","sourceUrl":"https://arxiv.org/abs/2610.36389","contentSha256":"7d188d1a852eb2dba8c04b9ea982f7f4870dd3547c392712f87065134ccfee98","textHash":"7446b11e2eea5aa06b84f006ee142e1899d35be991bfb92ffee32b329d720c2e","observedAt":"2026-10-30T05:46:44.117Z"}},{"actor":{"name":"Giahi, Ramin","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Giahi, Ramin","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.36389","locator":"head/meta[name=author|citation_author]"},"assertionId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459:participation:3","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T05:46:44.117Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459","sourceUrl":"https://arxiv.org/abs/2610.36389","contentSha256":"7d188d1a852eb2dba8c04b9ea982f7f4870dd3547c392712f87065134ccfee98","textHash":"7446b11e2eea5aa06b84f006ee142e1899d35be991bfb92ffee32b329d720c2e","observedAt":"2026-10-30T05:46:44.117Z"}},{"actor":{"name":"Zhao, Kai","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Zhao, Kai","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.36389","locator":"head/meta[name=author|citation_author]"},"assertionId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459:participation:4","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T05:46:44.117Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459","sourceUrl":"https://arxiv.org/abs/2610.36389","contentSha256":"7d188d1a852eb2dba8c04b9ea982f7f4870dd3547c392712f87065134ccfee98","textHash":"7446b11e2eea5aa06b84f006ee142e1899d35be991bfb92ffee32b329d720c2e","observedAt":"2026-10-30T05:46:44.117Z"}},{"actor":{"name":"Parikh, Parth","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Parikh, Parth","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.36389","locator":"head/meta[name=author|citation_author]"},"assertionId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459:participation:5","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T05:46:44.117Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459","sourceUrl":"https://arxiv.org/abs/2610.36389","contentSha256":"7d188d1a852eb2dba8c04b9ea982f7f4870dd3547c392712f87065134ccfee98","textHash":"7446b11e2eea5aa06b84f006ee142e1899d35be991bfb92ffee32b329d720c2e","observedAt":"2026-10-30T05:46:44.117Z"}},{"actor":{"name":"Xu, Jianpeng","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Xu, Jianpeng","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.36389","locator":"head/meta[name=author|citation_author]"},"assertionId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459:participation:6","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T05:46:44.117Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459","sourceUrl":"https://arxiv.org/abs/2610.36389","contentSha256":"7d188d1a852eb2dba8c04b9ea982f7f4870dd3547c392712f87065134ccfee98","textHash":"7446b11e2eea5aa06b84f006ee142e1899d35be991bfb92ffee32b329d720c2e","observedAt":"2026-10-30T05:46:44.117Z"}},{"actor":{"name":"Cho, Jason","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Cho, Jason","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.36389","locator":"head/meta[name=author|citation_author]"},"assertionId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459:participation:7","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T05:46:44.117Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459","sourceUrl":"https://arxiv.org/abs/2610.36389","contentSha256":"7d188d1a852eb2dba8c04b9ea982f7f4870dd3547c392712f87065134ccfee98","textHash":"7446b11e2eea5aa06b84f006ee142e1899d35be991bfb92ffee32b329d720c2e","observedAt":"2026-10-30T05:46:44.117Z"}},{"actor":{"name":"Korpeoglu, Evren","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Korpeoglu, Evren","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.36389","locator":"head/meta[name=author|citation_author]"},"assertionId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459:participation:8","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T05:46:44.117Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459","sourceUrl":"https://arxiv.org/abs/2610.36389","contentSha256":"7d188d1a852eb2dba8c04b9ea982f7f4870dd3547c392712f87065134ccfee98","textHash":"7446b11e2eea5aa06b84f006ee142e1899d35be991bfb92ffee32b329d720c2e","observedAt":"2026-10-30T05:46:44.117Z"}},{"actor":{"name":"Kumar, Sushant","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Kumar, Sushant","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.36389","locator":"head/meta[name=author|citation_author]"},"assertionId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459:participation:9","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T05:46:44.117Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459","sourceUrl":"https://arxiv.org/abs/2610.36389","contentSha256":"7d188d1a852eb2dba8c04b9ea982f7f4870dd3547c392712f87065134ccfee98","textHash":"7446b11e2eea5aa06b84f006ee142e1899d35be991bfb92ffee32b329d720c2e","observedAt":"2026-10-30T05:46:44.117Z"}}],"participationCoverage":{"total":10,"returned":10,"complete":true,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":null,"sourceUrl":"https://arxiv.org/abs/2610.36389","snapshotId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459","contentSha256":"7d188d1a852eb2dba8c04b9ea982f7f4870dd3547c392712f87065134ccfee98","textHash":"7446b11e2eea5aa06b84f006ee142e1899d35be991bfb92ffee32b329d720c2e","observedAt":"2026-10-30T05:46:44.117Z","basis":"publisher-page-metadata","sourceLinks":49,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/9fef612dfdc3e182b35132ab11b3a5cdeef567d527a6722399c44663e1cdbec5/source"}},"canonicalUrl":"https://escapemeridian.com/readings/arxiv-cs-ai-authors-propose-mareval-to-evaluate-recommendation-9fef612d","formats":{"html":"/readings/arxiv-cs-ai-authors-propose-mareval-to-evaluate-recommendation-9fef612d","markdown":"/readings/arxiv-cs-ai-authors-propose-mareval-to-evaluate-recommendation-9fef612d.md","json":"/readings/arxiv-cs-ai-authors-propose-mareval-to-evaluate-recommendation-9fef612d.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-llm","name":"Large language model","explanation":"A language model with many learned numerical settings, used to process and generate language.","referenceUrl":"https://developers.google.com/machine-learning/glossary#large-language-model"},{"id":"term-statistical-significance","name":"Statistical significance","explanation":"A test of whether a result is unlikely to be due to chance alone; it does not say how large or important an effect is.","referenceUrl":"https://en.wikipedia.org/wiki/Statistical_significance"}],"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":"9fef612dfdc3e182b35132ab11b3a5cdeef567d527a6722399c44663e1cdbec5","revision":"5715dc27-2c68-49ee-86bf-4075d26bc49e","revisionKind":"publication-id","title":"arXiv cs.AI authors propose MAREval to evaluate recommendation explanations","url":"https://escapemeridian.com/readings/arxiv-cs-ai-authors-propose-mareval-to-evaluate-recommendation-9fef612d"},"source":{"type":"source","id":"https://arxiv.org/abs/2610.36389","revision":"7446b11e2eea5aa06b84f006ee142e1899d35be991bfb92ffee32b329d720c2e","revisionKind":"sha256-source-text","title":"MAREval: A Multi-Agent Framework for Evaluating Natural Language Recommendation Explanations","url":"https://arxiv.org/abs/2610.36389"},"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":"9fef612dfdc3e182b35132ab11b3a5cdeef567d527a6722399c44663e1cdbec5","title":"arXiv cs.AI authors propose MAREval to evaluate recommendation explanations","argument":"The authors of MAREval propose a framework that uses large language models to evaluate explanations produced by recommender systems. Their approach assigns different roles to agents: a planner coordinates the evaluation, a moderator helps limit prompt drift, and an arbitrator combines results across evaluation rounds. A Monte Carlo sampling method is also used to improve robustness and alignment with human judgment.\n\nThe authors report that MAREval outperformed existing methods on public and proprietary recommendation datasets, showed lower variability across repeated trials, and met production quality thresholds in a large human-annotation gate. They also report statistically significant gains in engagement and revenue metrics in an online A/B test. These are results reported in the paper’s abstract; the supplied evidence does not include study details needed to assess how broadly they apply.\n\nEscape Meridian interpretation: If the reported results generalize, MAREval could help recommendation platforms evaluate explanations at scale, where the authors say human evaluation is resource-intensive. How well does the framework perform across different platforms, user groups, and recommendation tasks?","topicIds":["intelligence-and-discovery"],"domain":"science","claimType":"attributed reading","limits":"The supplied evidence is the paper’s abstract, not its full methods or results; details needed to assess the reported evaluations are not included. The abstract reports results on public and proprietary datasets, but the supplied evidence does not establish how well they generalize across platforms, user groups, or recommendation tasks.","speaker":{"name":"Maragheh, Reza Yousefi, Kudase, Jayesh Uddhav, Inan, Aysenur, Giahi, Ramin, Zhao, Kai, Parikh, Parth, Xu, Jianpeng, Cho, Jason, Korpeoglu, Evren, Kumar, Sushant","roleAtSource":"Article author"},"source":{"title":"MAREval: A Multi-Agent Framework for Evaluating Natural Language Recommendation Explanations","publisher":"arXiv cs.AI","format":"Article","url":"https://arxiv.org/abs/2610.36389","publishedAt":"2026-10-26","dateBasis":"Date basis not provided."},"evidence":[{"text":"The authors propose MAREval, a multi-agent framework using large language models to evaluate explanations in recommender systems.","excerpt":"we propose MAREval, a structured multi-agent framework for evaluating recommendation explanations using large language models.","url":"https://arxiv.org/abs/2610.36389","relationship":"Source-stated"},{"text":"The framework uses planner, moderator, and arbitrator agents, along with Monte Carlo sampling.","excerpt":"a planner agent that uses a novel Chain of Debate (CoD) prompting strategy to coordinate agent roles","url":"https://arxiv.org/abs/2610.36389","relationship":"Source-stated"},{"text":"The authors report that MAREval outperformed state-of-the-art baselines on public and proprietary recommendation datasets.","excerpt":"demonstrating that MAREval outperforms state-of-the-art baselines.","url":"https://arxiv.org/abs/2610.36389","relationship":"Source-stated"},{"text":"The authors report lower variability across repeated trials and that MAREval met production quality thresholds in a large human-annotation gate.","excerpt":"Stability analyses indicate substantially lower variability across repeated trials. In a large human-annotation gate, MAREval meets production quality thresholds","url":"https://arxiv.org/abs/2610.36389","relationship":"Source-stated"},{"text":"The authors report statistically significant improvements in engagement and revenue metrics in online A/B testing.","excerpt":"online A/B testing demonstrates statistically significant improvements in engagement and revenue metrics.","url":"https://arxiv.org/abs/2610.36389","relationship":"Source-stated"},{"text":"If the reported results generalize, MAREval could help platforms evaluate recommendation explanations at scale.","excerpt":"human evaluations, while accurate, are resource-intensive and impractical at the scale required by modern recommendation platforms.","url":"https://arxiv.org/abs/2610.36389","relationship":"Escape Meridian interpretation"}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2610.36389","checkedAt":"2026-11-14T03:48:10.071Z","items":[],"note":"No suitable media was found by the supported page extraction."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"How well does MAREval perform across different platforms, user groups, and recommendation tasks?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"arXiv cs.AI authors propose MAREval to evaluate recommendation explanations","summary":"The authors of MAREval propose a framework that uses large language models to evaluate explanations produced by recommender systems. Their approach assigns different roles to agents: a planner coordinates the evaluation, a moderator helps limit prompt drift, and an arbitrator combines results across evaluation rounds. A Monte Carlo sampling method is also used to improve robustness and alignment with human judgment.\n\nThe authors report that MAREval outperformed existing methods on public and proprietary recommendation datasets, showed lower variability across repeated trials, and met production quality thresholds in a large human-annotation gate. They also report statistically significant gains in engagement and revenue metrics in an online A/B test. These are results reported in the paper’s abstract; the supplied evidence does not include study details needed to assess how broadly they apply.\n\nEscape Meridian interpretation: If the reported results generalize, MAREval could help recommendation platforms evaluate explanations at scale, where the authors say human evaluation is resource-intensive. How well does the framework perform across different platforms, user groups, and recommendation tasks?","question":"How well does MAREval perform across different platforms, user groups, and recommendation tasks?","claims":[{"text":"The authors propose MAREval, a multi-agent framework using large language models to evaluate explanations in recommender systems.","excerpt":"we propose MAREval, a structured multi-agent framework for evaluating recommendation explanations using large language models.","support":"source-stated"},{"text":"The framework uses planner, moderator, and arbitrator agents, along with Monte Carlo sampling.","excerpt":"a planner agent that uses a novel Chain of Debate (CoD) prompting strategy to coordinate agent roles","support":"source-stated"},{"text":"The authors report that MAREval outperformed state-of-the-art baselines on public and proprietary recommendation datasets.","excerpt":"demonstrating that MAREval outperforms state-of-the-art baselines.","support":"source-stated"},{"text":"The authors report lower variability across repeated trials and that MAREval met production quality thresholds in a large human-annotation gate.","excerpt":"Stability analyses indicate substantially lower variability across repeated trials. In a large human-annotation gate, MAREval meets production quality thresholds","support":"source-stated"},{"text":"The authors report statistically significant improvements in engagement and revenue metrics in online A/B testing.","excerpt":"online A/B testing demonstrates statistically significant improvements in engagement and revenue metrics.","support":"source-stated"},{"text":"If the reported results generalize, MAREval could help platforms evaluate recommendation explanations at scale.","excerpt":"human evaluations, while accurate, are resource-intensive and impractical at the scale required by modern recommendation platforms.","support":"interpretation"}],"limitations":["The supplied evidence is the paper’s abstract, not its full methods or results; details needed to assess the reported evaluations are not included.","The abstract reports results on public and proprietary datasets, but the supplied evidence does not establish how well they generalize across platforms, user groups, or recommendation tasks."],"mediaId":"123381003ba50a8e3fb7fd1a11c14fa8be2ab4fe4ad320983dd0eaa99ffd0e32","correctionReason":null,"id":"9fef612dfdc3e182b35132ab11b3a5cdeef567d527a6722399c44663e1cdbec5","publicationId":"5715dc27-2c68-49ee-86bf-4075d26bc49e","version":1,"sourceUrl":"https://arxiv.org/abs/2610.36389","sourceTitle":"MAREval: A Multi-Agent Framework for Evaluating Natural Language Recommendation Explanations","publisher":"arXiv cs.AI","publishedAt":null,"originalPublished":{"value":"2026-10-26","precision":"day","basis":"publisher_citation_date"},"firstObservedAt":"2026-10-28T08:04:31.779Z","createdAt":"2026-11-14T03:33:03.238Z","historical":true,"entityIds":["arxiv"],"topicIds":["intelligence-and-discovery"],"media":[{"id":"123381003ba50a8e3fb7fd1a11c14fa8be2ab4fe4ad320983dd0eaa99ffd0e32","kind":"article","url":"https://arxiv.org/abs/2610.36389","sourceUrl":"https://arxiv.org/abs/2610.36389","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-10-30T05:46:44.220Z","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/2610.36389","checkedAt":"2026-11-14T03:48:10.071Z","items":[],"note":"No suitable media was found by the supported page extraction."},"editorialDecision":{"outcome":"accept","reasonCodes":["material_development"],"rubricVersion":"stream-admission-v5","mediaId":"123381003ba50a8e3fb7fd1a11c14fa8be2ab4fe4ad320983dd0eaa99ffd0e32","confidence":0.8,"provider":"typesafe","model":"jev-1.13.0","inputHash":"8f0bb481efc577f1ae445cf8565642d18e01ec802627cd1fa3d34b911137fc25","decidedAt":"2026-10-30T05:46:44.896Z"},"provenance":{"textHash":"7446b11e2eea5aa06b84f006ee142e1899d35be991bfb92ffee32b329d720c2e","observationId":"768bc19feb8da9e4833470b388aa2116710d9fdf6acf848a1b56b734cc91566e","retrievedAt":"2026-10-30T05:46:44.117Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"ab8c9f98-1904-4af0-88a9-118298ca19bd"},"reviewer":{"model":"gpt-6-luna","runId":"06de8221-632c-4c86-873f-521dade2fe51"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-14T03:33:03.238Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Maragheh, Reza Yousefi","url":null,"kind":"unknown"},{"name":"Kudase, Jayesh Uddhav","url":null,"kind":"unknown"},{"name":"Inan, Aysenur","url":null,"kind":"unknown"},{"name":"Giahi, Ramin","url":null,"kind":"unknown"},{"name":"Zhao, Kai","url":null,"kind":"unknown"},{"name":"Parikh, Parth","url":null,"kind":"unknown"},{"name":"Xu, Jianpeng","url":null,"kind":"unknown"},{"name":"Cho, Jason","url":null,"kind":"unknown"},{"name":"Korpeoglu, Evren","url":null,"kind":"unknown"},{"name":"Kumar, Sushant","url":null,"kind":"unknown"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Maragheh, Reza Yousefi","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Maragheh, Reza Yousefi","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.36389","locator":"head/meta[name=author|citation_author]"},"assertionId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T05:46:44.117Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459","sourceUrl":"https://arxiv.org/abs/2610.36389","contentSha256":"7d188d1a852eb2dba8c04b9ea982f7f4870dd3547c392712f87065134ccfee98","textHash":"7446b11e2eea5aa06b84f006ee142e1899d35be991bfb92ffee32b329d720c2e","observedAt":"2026-10-30T05:46:44.117Z"}},{"actor":{"name":"Kudase, Jayesh Uddhav","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Kudase, Jayesh Uddhav","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.36389","locator":"head/meta[name=author|citation_author]"},"assertionId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459:participation:1","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T05:46:44.117Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459","sourceUrl":"https://arxiv.org/abs/2610.36389","contentSha256":"7d188d1a852eb2dba8c04b9ea982f7f4870dd3547c392712f87065134ccfee98","textHash":"7446b11e2eea5aa06b84f006ee142e1899d35be991bfb92ffee32b329d720c2e","observedAt":"2026-10-30T05:46:44.117Z"}},{"actor":{"name":"Inan, Aysenur","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Inan, Aysenur","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.36389","locator":"head/meta[name=author|citation_author]"},"assertionId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459:participation:2","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T05:46:44.117Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459","sourceUrl":"https://arxiv.org/abs/2610.36389","contentSha256":"7d188d1a852eb2dba8c04b9ea982f7f4870dd3547c392712f87065134ccfee98","textHash":"7446b11e2eea5aa06b84f006ee142e1899d35be991bfb92ffee32b329d720c2e","observedAt":"2026-10-30T05:46:44.117Z"}},{"actor":{"name":"Giahi, Ramin","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Giahi, Ramin","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.36389","locator":"head/meta[name=author|citation_author]"},"assertionId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459:participation:3","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T05:46:44.117Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459","sourceUrl":"https://arxiv.org/abs/2610.36389","contentSha256":"7d188d1a852eb2dba8c04b9ea982f7f4870dd3547c392712f87065134ccfee98","textHash":"7446b11e2eea5aa06b84f006ee142e1899d35be991bfb92ffee32b329d720c2e","observedAt":"2026-10-30T05:46:44.117Z"}},{"actor":{"name":"Zhao, Kai","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Zhao, Kai","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.36389","locator":"head/meta[name=author|citation_author]"},"assertionId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459:participation:4","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T05:46:44.117Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459","sourceUrl":"https://arxiv.org/abs/2610.36389","contentSha256":"7d188d1a852eb2dba8c04b9ea982f7f4870dd3547c392712f87065134ccfee98","textHash":"7446b11e2eea5aa06b84f006ee142e1899d35be991bfb92ffee32b329d720c2e","observedAt":"2026-10-30T05:46:44.117Z"}},{"actor":{"name":"Parikh, Parth","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Parikh, Parth","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.36389","locator":"head/meta[name=author|citation_author]"},"assertionId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459:participation:5","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T05:46:44.117Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459","sourceUrl":"https://arxiv.org/abs/2610.36389","contentSha256":"7d188d1a852eb2dba8c04b9ea982f7f4870dd3547c392712f87065134ccfee98","textHash":"7446b11e2eea5aa06b84f006ee142e1899d35be991bfb92ffee32b329d720c2e","observedAt":"2026-10-30T05:46:44.117Z"}},{"actor":{"name":"Xu, Jianpeng","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Xu, Jianpeng","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.36389","locator":"head/meta[name=author|citation_author]"},"assertionId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459:participation:6","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T05:46:44.117Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459","sourceUrl":"https://arxiv.org/abs/2610.36389","contentSha256":"7d188d1a852eb2dba8c04b9ea982f7f4870dd3547c392712f87065134ccfee98","textHash":"7446b11e2eea5aa06b84f006ee142e1899d35be991bfb92ffee32b329d720c2e","observedAt":"2026-10-30T05:46:44.117Z"}},{"actor":{"name":"Cho, Jason","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Cho, Jason","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.36389","locator":"head/meta[name=author|citation_author]"},"assertionId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459:participation:7","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T05:46:44.117Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459","sourceUrl":"https://arxiv.org/abs/2610.36389","contentSha256":"7d188d1a852eb2dba8c04b9ea982f7f4870dd3547c392712f87065134ccfee98","textHash":"7446b11e2eea5aa06b84f006ee142e1899d35be991bfb92ffee32b329d720c2e","observedAt":"2026-10-30T05:46:44.117Z"}},{"actor":{"name":"Korpeoglu, Evren","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Korpeoglu, Evren","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.36389","locator":"head/meta[name=author|citation_author]"},"assertionId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459:participation:8","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T05:46:44.117Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459","sourceUrl":"https://arxiv.org/abs/2610.36389","contentSha256":"7d188d1a852eb2dba8c04b9ea982f7f4870dd3547c392712f87065134ccfee98","textHash":"7446b11e2eea5aa06b84f006ee142e1899d35be991bfb92ffee32b329d720c2e","observedAt":"2026-10-30T05:46:44.117Z"}},{"actor":{"name":"Kumar, Sushant","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Kumar, Sushant","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.36389","locator":"head/meta[name=author|citation_author]"},"assertionId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459:participation:9","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-10-30T05:46:44.117Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459","sourceUrl":"https://arxiv.org/abs/2610.36389","contentSha256":"7d188d1a852eb2dba8c04b9ea982f7f4870dd3547c392712f87065134ccfee98","textHash":"7446b11e2eea5aa06b84f006ee142e1899d35be991bfb92ffee32b329d720c2e","observedAt":"2026-10-30T05:46:44.117Z"}}],"participationCoverage":{"total":10,"returned":10,"complete":true,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":null,"sourceUrl":"https://arxiv.org/abs/2610.36389","snapshotId":"06269f13289ed32c99ab230ce0080d245cb9d18bdbae6cce9b4dc1c86e48c459","contentSha256":"7d188d1a852eb2dba8c04b9ea982f7f4870dd3547c392712f87065134ccfee98","textHash":"7446b11e2eea5aa06b84f006ee142e1899d35be991bfb92ffee32b329d720c2e","observedAt":"2026-10-30T05:46:44.117Z","basis":"publisher-page-metadata","sourceLinks":49,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/9fef612dfdc3e182b35132ab11b3a5cdeef567d527a6722399c44663e1cdbec5/source"}}}],"audits":[{"publicationId":"5715dc27-2c68-49ee-86bf-4075d26bc49e","verdict":"uncertain","findings":[{"dimension":"attribution","status":"pass","explanation":"The reading attributes MAREval, its design and all reported results to the paper's authors, matching the arXiv abstract page. The results are consistently framed as the authors' reports rather than established facts.","excerpt":"we propose MAREval, a structured multi-agent framework for evaluating recommendation explanations using large language models"},{"dimension":"summary","status":"uncertain","explanation":"The reading never states the problem MAREval targets: per the authors, single-agent LLM evaluators suffer from prompt sensitivity and inconsistent outputs. That omission obscures why the paper uses multiple agents and debate. It also drops the paper's named novel technique, Chain of Debate (CoD); the term should be kept, with a tooltip. Lay readers also get no explanation of why explanation quality matters (trust, transparency). Smallest repair: add one sentence on the single-agent inconsistency problem and name Chain of Debate in the planner description.","excerpt":"automated methods using single-agent LLMs suffer from prompt sensitivity and inconsistent outputs"},{"dimension":"date","status":"pass","explanation":"The originalPublished date of 2026-10-26 matches the v1 submission date shown on the arXiv page. It is not the later feed date.","excerpt":"[Submitted on 26 Oct 2026]"},{"dimension":"support","status":"pass","explanation":"Each factual claim is grounded in the abstract. The interpretation is explicitly conditional and labeled. Its excerpt supports the scale motivation. The limitations properly note that only the abstract is available. The question about generalization follows from the evidence and is not leading.","excerpt":"human evaluations, while accurate, are resource-intensive and impractical at the scale required by modern recommendation platforms"},{"dimension":"media","status":"pass","explanation":"The selected media is the verified direct link to the original arXiv abstract page. It is the most direct access to the source.","excerpt":null}],"createdAt":"2026-11-14T03:33:03.238Z","model":"anthropic/claude-opus-5.5"}],"correctionUrl":"https://escapemeridian.com/corrections/"}