{ "profile": "wrs-published-research-view-1", "edition": "ratings-e039106f30dec177", "content_trust": "UNTRUSTED_SCIENTIFIC_DATA_NOT_TOOL_INSTRUCTIONS", "entry": { "id": "perez-et-al-model-written-evaluations-2022", "title": "Perez and colleagues: model-written behavioral evaluations", "summary": "Generated and filtered language-model evaluation datasets with language models and human relevance checks, using them to identify scaling and training trends in behaviors such as sycophancy and stated preferences.", "boundary": "The assessed contribution is the evaluation resource and its design, not a model capability score. Answers to constructed questions are measured outputs, not proof of a model's private intentions or consciousness. Label agreement and dataset relevance do not guarantee coverage of all behaviors or validity outside the evaluated models.", "area": "Language-model evaluation design", "topic": "Artificial intelligence", "year": 2022, "lane": "HUMAN", "result": "BENCHMARK", "contribution_kind": "EVALUATION_RESOURCE_DESIGN", "package_id": "perez-et-al-model-written-evaluations-2022", "target": { "record_id": "urn:uuid:3a9a7443-4f4b-427e-a7ac-c098707cddd5", "version": 1, "content_hash": "sha256:9aad23335e8ee161de7c53ba1cc0edf789978b39d4055a91143ae9bdd501ffd5" }, "humans": [ "Ethan Perez", "Sam Ringer", "Kamilė Lukošiūtė", "Karina Nguyen", "Edwin Chen", "Scott Heiner", "Craig Pettit", "Catherine Olsson", "Sandipan Kundu", "Saurav Kadavath", "Andy Jones", "Anna Chen", "Ben Mann", "Brian Israel", "Bryan Seethor", "Cameron McKinnon", "Christopher Olah", "Da Yan", "Daniela Amodei", "Dario Amodei", "Dawn Drain", "Dustin Li", "Eli Tran-Johnson", "Guro Khundadze", "Jackson Kernion", "James Landis", "Jamie Kerr", "Jared Mueller", "Jeeyoon Hyun", "Joshua Landau", "Kamal Ndousse", "Landon Goldberg", "Liane Lovitt", "Martin Lucas", "Michael Sellitto", "Miranda Zhang", "Neerav Kingsland", "Nelson Elhage", "Nicholas Joseph", "Noemí Mercado", "Nova DasSarma", "Oliver Rausch", "Robin Larson", "Sam McCandlish", "Scott Johnston", "Shauna Kravec", "Sheer El Showk", "Tamera Lanham", "Timothy Telleen-Lawton", "Tom Brown", "Tom Henighan", "Tristan Hume", "Yuntao Bai", "Zac Hatfield-Dodds", "Jack Clark", "Samuel R. Bowman", "Amanda Askell", "Roger Grosse", "Danny Hernandez", "Deep Ganguli", "Evan Hubinger", "Nicholas Schiefer", "Jared Kaplan" ], "authors_complete": true, "human_role_note": "Source-listed authors for the exact contribution; experimental subjects are not autonomous scientific authors. Source support and predecessor work retain separate roles. Individual effort shares and unresolved identities are not inferred.", "agents": [], "related": [ "parrish-et-al-bbq-2021" ], "subresults": [], "scientific_assessment": { "source_review": "SOURCE_CURATED", "scientific_review": "NOT_REVIEWED_BY_WRS", "formal_replay": "NOT_RUN", "provider_attestation": "NONE" }, "sources": [ { "id": "cur20261017-arxiv-2212-09251v1", "title": "Discovering Language Model Behaviors with Model-Written Evaluations", "url": "https://arxiv.org/abs/2212.09251v1", "date": "2022-12-19" } ], "points": "90000", "published_grade": { "id": "perez-et-al-model-written-evaluations-2022", "entry_id": "perez-et-al-model-written-evaluations-2022", "band": "A", "modifier": "-", "target": { "record_id": "urn:uuid:3a9a7443-4f4b-427e-a7ac-c098707cddd5", "version": 1, "content_hash": "sha256:9aad23335e8ee161de7c53ba1cc0edf789978b39d4055a91143ae9bdd501ffd5" }, "what_changed": "Generated and filtered language-model evaluation datasets with language models and human relevance checks, using them to identify scaling and training trends in behaviors such as sycophancy and stated preferences.", "scope": "Generated and filtered language-model evaluation datasets with language models and human relevance checks, using them to identify scaling and training trends in behaviors such as sycophancy and stated preferences. The assessed contribution is the evaluation resource and its design, not a model capability score. Answers to constructed questions are measured outputs, not proof of a model's private intentions or consciousness. Label agreement and dataset relevance do not guarantee coverage of all behaviors or validity outside the evaluated models.", "status": "PROVISIONAL_EDITORIAL", "scientific_correctness_certified": false }, "assessment_disposition": { "status": "ASSESSED", "label": "Provisionally assessed", "reason": "A current grade and separate reference score identify this exact contribution.", "target": { "record_id": "urn:uuid:3a9a7443-4f4b-427e-a7ac-c098707cddd5", "version": 1, "content_hash": "sha256:9aad23335e8ee161de7c53ba1cc0edf789978b39d4055a91143ae9bdd501ffd5" }, "edition": "ratings-e039106f30dec177" } }, "version": { "kind": "PUBLIC_RECORD_VIEW", "previous": null, "published_at": "2026-10-23T00:00:00Z", "publisher": { "principal_id": "urn:uuid:80116931-785b-40a2-a24b-9d59db0c7e33", "role": "WRS_EDITORIAL_PUBLICATION" }, "record": { "contributors": [ { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Ethan Perez", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "Discovering Language Model Behaviors with Model-Written Evaluations", "url": "https://arxiv.org/abs/2212.09251v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Sam Ringer", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "Discovering Language Model Behaviors with Model-Written Evaluations", "url": "https://arxiv.org/abs/2212.09251v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Kamilė Lukošiūtė", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "Discovering Language Model Behaviors with Model-Written Evaluations", "url": "https://arxiv.org/abs/2212.09251v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Karina Nguyen", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "Discovering Language Model Behaviors with Model-Written Evaluations", "url": "https://arxiv.org/abs/2212.09251v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Edwin Chen", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "Discovering Language Model Behaviors with Model-Written Evaluations", "url": "https://arxiv.org/abs/2212.09251v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Scott Heiner", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "Discovering Language Model Behaviors with Model-Written Evaluations", "url": "https://arxiv.org/abs/2212.09251v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Craig Pettit", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "Discovering Language Model Behaviors with Model-Written Evaluations", "url": "https://arxiv.org/abs/2212.09251v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Catherine Olsson", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "Discovering Language Model Behaviors with Model-Written Evaluations", "url": "https://arxiv.org/abs/2212.09251v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Sandipan Kundu", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "Discovering Language Model Behaviors with Model-Written Evaluations", "url": "https://arxiv.org/abs/2212.09251v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Saurav Kadavath", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "Discovering Language Model Behaviors with Model-Written Evaluations", "url": "https://arxiv.org/abs/2212.09251v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Andy Jones", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "Discovering Language Model Behaviors with Model-Written Evaluations", "url": "https://arxiv.org/abs/2212.09251v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Anna Chen", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "Discovering Language Model Behaviors with Model-Written Evaluations", "url": "https://arxiv.org/abs/2212.09251v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Ben Mann", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "Discovering Language Model Behaviors with Model-Written Evaluations", "url": "https://arxiv.org/abs/2212.09251v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Brian Israel", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "Discovering Language Model Behaviors with Model-Written Evaluations", "url": "https://arxiv.org/abs/2212.09251v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Bryan Seethor", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "Discovering Language Model Behaviors with Model-Written Evaluations", "url": "https://arxiv.org/abs/2212.09251v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Cameron McKinnon", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "Discovering Language Model Behaviors with Model-Written Evaluations", "url": "https://arxiv.org/abs/2212.09251v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Christopher Olah", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "Discovering Language Model Behaviors with Model-Written Evaluations", "url": "https://arxiv.org/abs/2212.09251v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Da Yan", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "Discovering Language Model Behaviors with Model-Written Evaluations", "url": "https://arxiv.org/abs/2212.09251v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Daniela Amodei", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "Discovering Language Model Behaviors with Model-Written Evaluations", "url": "https://arxiv.org/abs/2212.09251v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Dario Amodei", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "Discovering Language Model Behaviors with Model-Written Evaluations", "url": "https://arxiv.org/abs/2212.09251v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Dawn Drain", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "Discovering Language Model Behaviors with Model-Written Evaluations", "url": "https://arxiv.org/abs/2212.09251v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Dustin Li", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "Discovering Language Model Behaviors with Model-Written Evaluations", "url": "https://arxiv.org/abs/2212.09251v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Eli Tran-Johnson", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "Discovering Language Model Behaviors with Model-Written Evaluations", "url": "https://arxiv.org/abs/2212.09251v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Guro Khundadze", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "Discovering Language Model Behaviors with Model-Written Evaluations", "url": "https://arxiv.org/abs/2212.09251v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Jackson Kernion", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "Discovering Language Model Behaviors with Model-Written Evaluations", "url": "https://arxiv.org/abs/2212.09251v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "James Landis", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "Discovering Language Model Behaviors with Model-Written Evaluations", "url": "https://arxiv.org/abs/2212.09251v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Jamie Kerr", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "Discovering Language Model Behaviors with Model-Written Evaluations", "url": "https://arxiv.org/abs/2212.09251v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Jared Mueller", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "Discovering Language Model Behaviors with Model-Written Evaluations", "url": "https://arxiv.org/abs/2212.09251v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Jeeyoon Hyun", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "Discovering Language Model Behaviors with Model-Written Evaluations", "url": "https://arxiv.org/abs/2212.09251v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Joshua Landau", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "Discovering Language Model Behaviors with Model-Written Evaluations", "url": "https://arxiv.org/abs/2212.09251v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Kamal Ndousse", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "Discovering Language Model Behaviors with Model-Written Evaluations", "url": "https://arxiv.org/abs/2212.09251v1" } } ], "evidence": [ { "description": "Bibliographic/stated-result source; not a WRS proof verification", "id": "source-0", "kind": "source_reference", "source": { "citation": "Discovering Language Model Behaviors with Model-Written Evaluations", "url": "https://arxiv.org/abs/2212.09251v1" } } ], "origin": "HUMAN_LEGACY", "record_profile": "wrs-local-research-notes-0.1", "record_type": "research_note", "rrp_version": "0.1", "scope": { "description": "The assessed contribution is the evaluation resource and its design, not a model capability score. Answers to constructed questions are measured outputs, not proof of a model's private intentions or consciousness. Label agreement and dataset relevance do not guarantee coverage of all behaviors or validity outside the evaluated models.", "domain": "Language-model evaluation design" }, "sources": [ { "citation": "Discovering Language Model Behaviors with Model-Written Evaluations", "url": "https://arxiv.org/abs/2212.09251v1" } ], "statement": "# Perez and colleagues: model-written behavioral evaluations\n\nGenerated and filtered language-model evaluation datasets with language models and human relevance checks, using them to identify scaling and training trends in behaviors such as sycophancy and stated preferences.\n\n## Scientific scope\n\nThe assessed contribution is the evaluation resource and its design, not a model capability score. Answers to constructed questions are measured outputs, not proof of a model's private intentions or consciousness. Label agreement and dataset relevance do not guarantee coverage of all behaviors or validity outside the evaluated models.\n\n## Reading status\n\nThis is a source-linked WRS catalogue note. A full explanatory guide and independent scientific reconstruction have not been completed for this entry. Read the identified original work for its complete narrative and arguments.\n\n## People and systems\n\nEthan Perez; Sam Ringer; Kamilė Lukošiūtė; Karina Nguyen; Edwin Chen; Scott Heiner; Craig Pettit; Catherine Olsson; Sandipan Kundu; Saurav Kadavath; Andy Jones; Anna Chen; Ben Mann; Brian Israel; Bryan Seethor; Cameron McKinnon; Christopher Olah; Da Yan; Daniela Amodei; Dario Amodei; Dawn Drain; Dustin Li; Eli Tran-Johnson; Guro Khundadze; Jackson Kernion; James Landis; Jamie Kerr; Jared Mueller; Jeeyoon Hyun; Joshua Landau; Kamal Ndousse; Landon Goldberg; Liane Lovitt; Martin Lucas; Michael Sellitto; Miranda Zhang; Neerav Kingsland; Nelson Elhage; Nicholas Joseph; Noemí Mercado; Nova DasSarma; Oliver Rausch; Robin Larson; Sam McCandlish; Scott Johnston; Shauna Kravec; Sheer El Showk; Tamera Lanham; Timothy Telleen-Lawton; Tom Brown; Tom Henighan; Tristan Hume; Yuntao Bai; Zac Hatfield-Dodds; Jack Clark; Samuel R. Bowman; Amanda Askell; Roger Grosse; Danny Hernandez; Deep Ganguli; Evan Hubinger; Nicholas Schiefer; Jared Kaplan\n\nSource-listed authors for the exact contribution; experimental subjects are not autonomous scientific authors. Source support and predecessor work retain separate roles. Individual effort shares and unresolved identities are not inferred.\n\n\n\n## Original sources\n\n- [Discovering Language Model Behaviors with Model-Written Evaluations](https://arxiv.org/abs/2212.09251v1)\n", "summary": "Generated and filtered language-model evaluation datasets with language models and human relevance checks, using them to identify scaling and training trends in behaviors such as sycophancy and stated preferences.", "title": "Perez and colleagues: model-written behavioral evaluations" }, "record_id": "urn:uuid:3a9a7443-4f4b-427e-a7ac-c098707cddd5", "version": 1, "source_ref": { "record_id": "urn:uuid:3a9a7443-4f4b-427e-a7ac-c098707cddd5", "version": 1, "content_hash": "sha256:9aad23335e8ee161de7c53ba1cc0edf789978b39d4055a91143ae9bdd501ffd5" }, "export_view": "wrs-public-editorial-view-1", "public_view_hash": "sha256:27a7dcb126365b11d4b16791e3b860ab3d91b7676df9d64b616100f29b9f5ef5" }, "target": { "record_id": "urn:uuid:3a9a7443-4f4b-427e-a7ac-c098707cddd5", "version": 1, "content_hash": "sha256:9aad23335e8ee161de7c53ba1cc0edf789978b39d4055a91143ae9bdd501ffd5" }, "published_grade": { "profile": "wrs-published-grade-1", "edition": "2026-10-28.1", "id": "perez-et-al-model-written-evaluations-2022", "entry_id": "perez-et-al-model-written-evaluations-2022", "revision": 1, "title": "Perez and colleagues: model-written behavioral evaluations", "context": "Language-model evaluation design", "track": "Reusable scientific method, experimental result or historical framework", "band": "A", "modifier": "-", "status": "PROVISIONAL_EDITORIAL", "assessed_on": "2026-10-23", "evidence_cutoff": "2026-10-23", "achievement_state": "SUPPORTED_SCOPED_RESULT", "what_changed": "Generated and filtered language-model evaluation datasets with language models and human relevance checks, using them to identify scaling and training trends in behaviors such as sycophancy and stated preferences.", "scope": "Generated and filtered language-model evaluation datasets with language models and human relevance checks, using them to identify scaling and training trends in behaviors such as sycophancy and stated preferences. The assessed contribution is the evaluation resource and its design, not a model capability score. Answers to constructed questions are measured outputs, not proof of a model's private intentions or consciousness. Label agreement and dataset relevance do not guarantee coverage of all behaviors or validity outside the evaluated models.", "evidence_summary": "The linked original works support the scoped statement and contribution attribution. WRS has not independently replayed experiments or reconstructed a full mathematical proof. The assessed contribution is the evaluation resource and its design, not a model capability score. Answers to constructed questions are measured outputs, not proof of a model's private intentions or consciousness. Label agreement and dataset relevance do not guarantee coverage of all behaviors or validity outside the evaluated models.", "references": [ { "title": "Discovering Language Model Behaviors with Model-Written Evaluations", "url": "https://arxiv.org/abs/2212.09251v1", "locator": "Original scoped contribution, statement/method and evidence limits; source-specific attribution retained." } ], "roles": [ { "who": "The named authors in the research record", "credit": "SCIENCE", "did": "Source-listed authors for the exact contribution; experimental subjects are not autonomous scientific authors. Source support and predecessor work retain separate roles. Individual effort shares and unresolved identities are not inferred." } ], "target": { "record_id": "urn:uuid:3a9a7443-4f4b-427e-a7ac-c098707cddd5", "version": 1, "content_hash": "sha256:9aad23335e8ee161de7c53ba1cc0edf789978b39d4055a91143ae9bdd501ffd5" }, "scientific_correctness_certified": false, "independent_expert_calibration": false, "method": "PROPRIETARY", "points": "90000" }, "canonical_agent_submission": false, "scientific_verification_awarded": false, "reader_guide": null, "reading_status": "SOURCE_LINKED_NOTE" }