{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:Y2TKICKAK2YGTW5XNRGWH7IWDA","short_pith_number":"pith:Y2TKICKA","schema_version":"1.0","canonical_sha256":"c6a6a4094056b069dbb76c4d63fd16180bb23430e86c520fb68e59ccd177eefb","source":{"kind":"arxiv","id":"2407.08639","version":2},"attestation_state":"computed","paper":{"title":"$\\beta$-DPO: Direct Preference Optimization with Dynamic $\\beta$","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.AI","authors_text":"Bolin Ding, Jiancan Wu, Jinyang Gao, Junkang Wu, Xiangnan He, Xiang Wang, Yuexiang Xie, Zhengyi Yang","submitted_at":"2024-07-11T16:21:18Z","abstract_excerpt":"Direct Preference Optimization (DPO) has emerged as a compelling approach for training Large Language Models (LLMs) to adhere to human preferences. However, the performance of DPO is sensitive to the fine-tuning of its trade-off parameter $\\beta$, as well as to the quality of the preference data. We analyze the impact of $\\beta$ and data quality on DPO, uncovering that optimal $\\beta$ values vary with the informativeness of pairwise data. Addressing the limitations of static $\\beta$ values, we introduce a novel framework that dynamically calibrates $\\beta$ at the batch level, informed by data "},"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":"2407.08639","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2024-07-11T16:21:18Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"814b14d176b816fa00061d31d6915ee586fda73c282b601b3b9c420944b0da0e","abstract_canon_sha256":"2249997031ee26ed620b230c128fe30272dc9759a113b2ffbac696922fbb7850"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:19:30.019452Z","signature_b64":"NPM6Mdcqx0s1BV639z0rJPSz/AywPmfSXesIdB19RhtxdXHFtJOHp5PRMPKeSGLz5gAP7JHHtwmdMBealajfAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c6a6a4094056b069dbb76c4d63fd16180bb23430e86c520fb68e59ccd177eefb","last_reissued_at":"2026-07-05T09:19:30.018970Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:19:30.018970Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"$\\beta$-DPO: Direct Preference Optimization with Dynamic $\\beta$","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.AI","authors_text":"Bolin Ding, Jiancan Wu, Jinyang Gao, Junkang Wu, Xiangnan He, Xiang Wang, Yuexiang Xie, Zhengyi Yang","submitted_at":"2024-07-11T16:21:18Z","abstract_excerpt":"Direct Preference Optimization (DPO) has emerged as a compelling approach for training Large Language Models (LLMs) to adhere to human preferences. However, the performance of DPO is sensitive to the fine-tuning of its trade-off parameter $\\beta$, as well as to the quality of the preference data. We analyze the impact of $\\beta$ and data quality on DPO, uncovering that optimal $\\beta$ values vary with the informativeness of pairwise data. Addressing the limitations of static $\\beta$ values, we introduce a novel framework that dynamically calibrates $\\beta$ at the batch level, informed by data "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2407.08639","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/2407.08639/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":"2407.08639","created_at":"2026-07-05T09:19:30.019027+00:00"},{"alias_kind":"arxiv_version","alias_value":"2407.08639v2","created_at":"2026-07-05T09:19:30.019027+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2407.08639","created_at":"2026-07-05T09:19:30.019027+00:00"},{"alias_kind":"pith_short_12","alias_value":"Y2TKICKAK2YG","created_at":"2026-07-05T09:19:30.019027+00:00"},{"alias_kind":"pith_short_16","alias_value":"Y2TKICKAK2YGTW5X","created_at":"2026-07-05T09:19:30.019027+00:00"},{"alias_kind":"pith_short_8","alias_value":"Y2TKICKA","created_at":"2026-07-05T09:19:30.019027+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.04539","citing_title":"RLearner-LLM: Balancing Logical Grounding and Fluency in Large Language Models via Hybrid Direct Preference Optimization","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2605.04539","citing_title":"RLearner-LLM: Balancing Logical Grounding and Fluency in Large Language Models via Hybrid Direct Preference Optimization","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2605.04539","citing_title":"RLearner-LLM: Balancing Logical Grounding and Fluency in Large Language Models via Hybrid Direct Preference Optimization","ref_index":9,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/Y2TKICKAK2YGTW5XNRGWH7IWDA","json":"https://pith.science/pith/Y2TKICKAK2YGTW5XNRGWH7IWDA.json","graph_json":"https://pith.science/api/pith-number/Y2TKICKAK2YGTW5XNRGWH7IWDA/graph.json","events_json":"https://pith.science/api/pith-number/Y2TKICKAK2YGTW5XNRGWH7IWDA/events.json","paper":"https://pith.science/paper/Y2TKICKA"},"agent_actions":{"view_html":"https://pith.science/pith/Y2TKICKAK2YGTW5XNRGWH7IWDA","download_json":"https://pith.science/pith/Y2TKICKAK2YGTW5XNRGWH7IWDA.json","view_paper":"https://pith.science/paper/Y2TKICKA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2407.08639&json=true","fetch_graph":"https://pith.science/api/pith-number/Y2TKICKAK2YGTW5XNRGWH7IWDA/graph.json","fetch_events":"https://pith.science/api/pith-number/Y2TKICKAK2YGTW5XNRGWH7IWDA/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/Y2TKICKAK2YGTW5XNRGWH7IWDA/action/timestamp_anchor","attest_storage":"https://pith.science/pith/Y2TKICKAK2YGTW5XNRGWH7IWDA/action/storage_attestation","attest_author":"https://pith.science/pith/Y2TKICKAK2YGTW5XNRGWH7IWDA/action/author_attestation","sign_citation":"https://pith.science/pith/Y2TKICKAK2YGTW5XNRGWH7IWDA/action/citation_signature","submit_replication":"https://pith.science/pith/Y2TKICKAK2YGTW5XNRGWH7IWDA/action/replication_record"}},"created_at":"2026-07-05T09:19:30.019027+00:00","updated_at":"2026-07-05T09:19:30.019027+00:00"}