{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:T2BHHEGJGAJJ7XE4NZMXUFXOLO","short_pith_number":"pith:T2BHHEGJ","schema_version":"1.0","canonical_sha256":"9e827390c930129fdc9c6e597a16ee5b94715449889977961d5a5b600f2553ad","source":{"kind":"arxiv","id":"2404.04102","version":2},"attestation_state":"computed","paper":{"title":"ROPO: Robust Preference Optimization for Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Chao Chen, Feng Wu, Jieping Ye, Jie Wang, Shuang Qiu, Xize Liang, Yue Wu, Zhihang Fu, Zhihao Shi","submitted_at":"2024-04-05T13:58:51Z","abstract_excerpt":"Preference alignment is pivotal for empowering large language models (LLMs) to generate helpful and harmless responses. However, the performance of preference alignment is highly sensitive to the prevalent noise in the preference data. Recent efforts for this problem either marginally alleviate the impact of noise without the ability to actually reduce its presence, or rely on costly teacher LLMs prone to reward misgeneralization. To address these challenges, we propose the RObust Preference Optimization (ROPO) framework, an iterative alignment approach that integrates noise-tolerance and filt"},"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":"2404.04102","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-04-05T13:58:51Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"b3508be7778c5fea26115096fccfe93f6f1638b09f245ae7064793a06a90f6bd","abstract_canon_sha256":"d8b68160dcd00e839814dc0210a64940ea4b091153659781b65b0906f3e6a80d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:24:04.294487Z","signature_b64":"hfPbsgr20SPcFX2SnO/mLxJZ8LDGnvkLJC55hPlg5kqdSWnsHGdurnAzfrrUw8CBJ3V48HKiCL+Qr5kMwWpRDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9e827390c930129fdc9c6e597a16ee5b94715449889977961d5a5b600f2553ad","last_reissued_at":"2026-07-05T08:24:04.293975Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:24:04.293975Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"ROPO: Robust Preference Optimization for Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Chao Chen, Feng Wu, Jieping Ye, Jie Wang, Shuang Qiu, Xize Liang, Yue Wu, Zhihang Fu, Zhihao Shi","submitted_at":"2024-04-05T13:58:51Z","abstract_excerpt":"Preference alignment is pivotal for empowering large language models (LLMs) to generate helpful and harmless responses. However, the performance of preference alignment is highly sensitive to the prevalent noise in the preference data. Recent efforts for this problem either marginally alleviate the impact of noise without the ability to actually reduce its presence, or rely on costly teacher LLMs prone to reward misgeneralization. To address these challenges, we propose the RObust Preference Optimization (ROPO) framework, an iterative alignment approach that integrates noise-tolerance and filt"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2404.04102","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/2404.04102/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":"2404.04102","created_at":"2026-07-05T08:24:04.294041+00:00"},{"alias_kind":"arxiv_version","alias_value":"2404.04102v2","created_at":"2026-07-05T08:24:04.294041+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2404.04102","created_at":"2026-07-05T08:24:04.294041+00:00"},{"alias_kind":"pith_short_12","alias_value":"T2BHHEGJGAJJ","created_at":"2026-07-05T08:24:04.294041+00:00"},{"alias_kind":"pith_short_16","alias_value":"T2BHHEGJGAJJ7XE4","created_at":"2026-07-05T08:24:04.294041+00:00"},{"alias_kind":"pith_short_8","alias_value":"T2BHHEGJ","created_at":"2026-07-05T08:24:04.294041+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2502.06387","citing_title":"How Humans Help LLMs: Assessing and Incentivizing Human Preference Annotators","ref_index":64,"is_internal_anchor":false},{"citing_arxiv_id":"2505.19134","citing_title":"Incentivizing High-Quality Human Annotations with Golden Questions","ref_index":28,"is_internal_anchor":false},{"citing_arxiv_id":"2510.13830","citing_title":"Users as Annotators: LLM Preference Learning from Comparison Mode","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06987","citing_title":"Response Time Enhances Alignment with Heterogeneous Preferences","ref_index":172,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/T2BHHEGJGAJJ7XE4NZMXUFXOLO","json":"https://pith.science/pith/T2BHHEGJGAJJ7XE4NZMXUFXOLO.json","graph_json":"https://pith.science/api/pith-number/T2BHHEGJGAJJ7XE4NZMXUFXOLO/graph.json","events_json":"https://pith.science/api/pith-number/T2BHHEGJGAJJ7XE4NZMXUFXOLO/events.json","paper":"https://pith.science/paper/T2BHHEGJ"},"agent_actions":{"view_html":"https://pith.science/pith/T2BHHEGJGAJJ7XE4NZMXUFXOLO","download_json":"https://pith.science/pith/T2BHHEGJGAJJ7XE4NZMXUFXOLO.json","view_paper":"https://pith.science/paper/T2BHHEGJ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2404.04102&json=true","fetch_graph":"https://pith.science/api/pith-number/T2BHHEGJGAJJ7XE4NZMXUFXOLO/graph.json","fetch_events":"https://pith.science/api/pith-number/T2BHHEGJGAJJ7XE4NZMXUFXOLO/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/T2BHHEGJGAJJ7XE4NZMXUFXOLO/action/timestamp_anchor","attest_storage":"https://pith.science/pith/T2BHHEGJGAJJ7XE4NZMXUFXOLO/action/storage_attestation","attest_author":"https://pith.science/pith/T2BHHEGJGAJJ7XE4NZMXUFXOLO/action/author_attestation","sign_citation":"https://pith.science/pith/T2BHHEGJGAJJ7XE4NZMXUFXOLO/action/citation_signature","submit_replication":"https://pith.science/pith/T2BHHEGJGAJJ7XE4NZMXUFXOLO/action/replication_record"}},"created_at":"2026-07-05T08:24:04.294041+00:00","updated_at":"2026-07-05T08:24:04.294041+00:00"}