{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:TJD4ZQGSBKCRB32SLHQRUGDKJO","short_pith_number":"pith:TJD4ZQGS","schema_version":"1.0","canonical_sha256":"9a47ccc0d20a8510ef5259e11a186a4b9b1fc19c073306c5763f7528d403961b","source":{"kind":"arxiv","id":"2312.16682","version":2},"attestation_state":"computed","paper":{"title":"Some things are more CRINGE than others: Iterative Preference Optimization with the Pairwise Cringe Loss","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Andrew Lee, Jason Weston, Jing Xu, Sainbayar Sukhbaatar","submitted_at":"2023-12-27T18:53:09Z","abstract_excerpt":"Practitioners commonly align large language models using pairwise preferences, i.e., given labels of the type response A is preferred to response B for a given input. Perhaps less commonly, methods have also been developed for binary feedback, i.e. training models given labels of type response A is good or bad. We show how an existing performant binary feedback method, the Cringe Loss (Adolphs et al., 2022), can be generalized to the pairwise preference setting using a simple soft margin extension. Pairwise Cringe Loss is straightforward to implement and efficient to train, and we find it outp"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2312.16682","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2023-12-27T18:53:09Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"70ce8c02607e37fa7013f5824e95e60c6bbb921c726867a704d519efc7fdc6b5","abstract_canon_sha256":"cb43e9394dc2478bebc3d0e8f64e426d7edaf5ca101c9c5c904566f12ea0d7a0"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:10:59.829656Z","signature_b64":"ae6YBtrJBu1N+K/CI+sfuz1Z1BfAZMwg23GQu4gyGMNkELHk6YrAGk7WLsw7ksg5lhagXb7Ju3H0Wr/mlFK+AQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9a47ccc0d20a8510ef5259e11a186a4b9b1fc19c073306c5763f7528d403961b","last_reissued_at":"2026-07-05T08:10:59.829100Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:10:59.829100Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Some things are more CRINGE than others: Iterative Preference Optimization with the Pairwise Cringe Loss","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Andrew Lee, Jason Weston, Jing Xu, Sainbayar Sukhbaatar","submitted_at":"2023-12-27T18:53:09Z","abstract_excerpt":"Practitioners commonly align large language models using pairwise preferences, i.e., given labels of the type response A is preferred to response B for a given input. Perhaps less commonly, methods have also been developed for binary feedback, i.e. training models given labels of type response A is good or bad. We show how an existing performant binary feedback method, the Cringe Loss (Adolphs et al., 2022), can be generalized to the pairwise preference setting using a simple soft margin extension. Pairwise Cringe Loss is straightforward to implement and efficient to train, and we find it outp"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2312.16682","kind":"arxiv","version":2},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2312.16682/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2312.16682","created_at":"2026-07-05T08:10:59.829159+00:00"},{"alias_kind":"arxiv_version","alias_value":"2312.16682v2","created_at":"2026-07-05T08:10:59.829159+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2312.16682","created_at":"2026-07-05T08:10:59.829159+00:00"},{"alias_kind":"pith_short_12","alias_value":"TJD4ZQGSBKCR","created_at":"2026-07-05T08:10:59.829159+00:00"},{"alias_kind":"pith_short_16","alias_value":"TJD4ZQGSBKCRB32S","created_at":"2026-07-05T08:10:59.829159+00:00"},{"alias_kind":"pith_short_8","alias_value":"TJD4ZQGS","created_at":"2026-07-05T08:10:59.829159+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":7,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.21090","citing_title":"Self-Improvement Can Self-Regress: The Rise-and-Collapse Failure Mode of LLM Self-Training","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2607.01470","citing_title":"World Feedback for Clinical Agents: Diagnosing RL in FHIR Environments","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2408.15339","citing_title":"UNA: A Unified Supervised Framework for Efficient LLM Alignment Across Feedback Types","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2509.20265","citing_title":"Failure Modes of Maximum Entropy RLHF","ref_index":56,"is_internal_anchor":false},{"citing_arxiv_id":"2605.15012","citing_title":"Boosting Reinforcement Learning with Verifiable Rewards via Randomly Selected Few-Shot Guidance","ref_index":89,"is_internal_anchor":false},{"citing_arxiv_id":"2401.10020","citing_title":"Self-Rewarding Language Models","ref_index":121,"is_internal_anchor":false},{"citing_arxiv_id":"2604.17543","citing_title":"PoliLegalLM: A Technical Report on a Large Language Model for Political and Legal Affairs","ref_index":40,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/TJD4ZQGSBKCRB32SLHQRUGDKJO","json":"https://pith.science/pith/TJD4ZQGSBKCRB32SLHQRUGDKJO.json","graph_json":"https://pith.science/api/pith-number/TJD4ZQGSBKCRB32SLHQRUGDKJO/graph.json","events_json":"https://pith.science/api/pith-number/TJD4ZQGSBKCRB32SLHQRUGDKJO/events.json","paper":"https://pith.science/paper/TJD4ZQGS"},"agent_actions":{"view_html":"https://pith.science/pith/TJD4ZQGSBKCRB32SLHQRUGDKJO","download_json":"https://pith.science/pith/TJD4ZQGSBKCRB32SLHQRUGDKJO.json","view_paper":"https://pith.science/paper/TJD4ZQGS","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2312.16682&json=true","fetch_graph":"https://pith.science/api/pith-number/TJD4ZQGSBKCRB32SLHQRUGDKJO/graph.json","fetch_events":"https://pith.science/api/pith-number/TJD4ZQGSBKCRB32SLHQRUGDKJO/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/TJD4ZQGSBKCRB32SLHQRUGDKJO/action/timestamp_anchor","attest_storage":"https://pith.science/pith/TJD4ZQGSBKCRB32SLHQRUGDKJO/action/storage_attestation","attest_author":"https://pith.science/pith/TJD4ZQGSBKCRB32SLHQRUGDKJO/action/author_attestation","sign_citation":"https://pith.science/pith/TJD4ZQGSBKCRB32SLHQRUGDKJO/action/citation_signature","submit_replication":"https://pith.science/pith/TJD4ZQGSBKCRB32SLHQRUGDKJO/action/replication_record"}},"created_at":"2026-07-05T08:10:59.829159+00:00","updated_at":"2026-07-05T08:10:59.829159+00:00"}