{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:WWVJAGVIL4CHPE247U25J7GJIB","short_pith_number":"pith:WWVJAGVI","schema_version":"1.0","canonical_sha256":"b5aa901aa85f0477935cfd35d4fcc940543106a43a13f3d3fbbc3cf2f7969161","source":{"kind":"arxiv","id":"2407.13928","version":1},"attestation_state":"computed","paper":{"title":"BiasDPO: Mitigating Bias in Language Models through Direct Preference Optimization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Ahmed Allam","submitted_at":"2024-07-18T22:32:20Z","abstract_excerpt":"Large Language Models (LLMs) have become pivotal in advancing natural language processing, yet their potential to perpetuate biases poses significant concerns. This paper introduces a new framework employing Direct Preference Optimization (DPO) to mitigate gender, racial, and religious biases in LLM-generated English text. By developing a loss function that favors less biased over biased completions, our approach cultivates a preference for respectful and non-discriminatory language in LLMs. We also contribute a manually designed dataset for training LLMs to recognize and correct biases. This "},"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":"2407.13928","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-07-18T22:32:20Z","cross_cats_sorted":[],"title_canon_sha256":"ad9e74c92e5e49315f8971292f7193803f21d901f84d030d11ce5e5f30c32ec5","abstract_canon_sha256":"430e2f76597f512134091f9025cf89f62ad2d42ba9c4a59795f42482983b6cec"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:45:55.655097Z","signature_b64":"22skYYtVX1csicC1driuRrTqYFmrvxccaM9+CdQpLvdKkKNZjwy9USV9/8sfcRk8Y91nfJWF/VRMf0Qh7PPgBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b5aa901aa85f0477935cfd35d4fcc940543106a43a13f3d3fbbc3cf2f7969161","last_reissued_at":"2026-07-05T08:45:55.654559Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:45:55.654559Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"BiasDPO: Mitigating Bias in Language Models through Direct Preference Optimization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Ahmed Allam","submitted_at":"2024-07-18T22:32:20Z","abstract_excerpt":"Large Language Models (LLMs) have become pivotal in advancing natural language processing, yet their potential to perpetuate biases poses significant concerns. This paper introduces a new framework employing Direct Preference Optimization (DPO) to mitigate gender, racial, and religious biases in LLM-generated English text. By developing a loss function that favors less biased over biased completions, our approach cultivates a preference for respectful and non-discriminatory language in LLMs. We also contribute a manually designed dataset for training LLMs to recognize and correct biases. This "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2407.13928","kind":"arxiv","version":1},"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/2407.13928/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":"2407.13928","created_at":"2026-07-05T08:45:55.654617+00:00"},{"alias_kind":"arxiv_version","alias_value":"2407.13928v1","created_at":"2026-07-05T08:45:55.654617+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2407.13928","created_at":"2026-07-05T08:45:55.654617+00:00"},{"alias_kind":"pith_short_12","alias_value":"WWVJAGVIL4CH","created_at":"2026-07-05T08:45:55.654617+00:00"},{"alias_kind":"pith_short_16","alias_value":"WWVJAGVIL4CHPE24","created_at":"2026-07-05T08:45:55.654617+00:00"},{"alias_kind":"pith_short_8","alias_value":"WWVJAGVI","created_at":"2026-07-05T08:45:55.654617+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.02585","citing_title":"Mitigating LLM biases toward spurious social contexts using direct preference optimization","ref_index":1,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/WWVJAGVIL4CHPE247U25J7GJIB","json":"https://pith.science/pith/WWVJAGVIL4CHPE247U25J7GJIB.json","graph_json":"https://pith.science/api/pith-number/WWVJAGVIL4CHPE247U25J7GJIB/graph.json","events_json":"https://pith.science/api/pith-number/WWVJAGVIL4CHPE247U25J7GJIB/events.json","paper":"https://pith.science/paper/WWVJAGVI"},"agent_actions":{"view_html":"https://pith.science/pith/WWVJAGVIL4CHPE247U25J7GJIB","download_json":"https://pith.science/pith/WWVJAGVIL4CHPE247U25J7GJIB.json","view_paper":"https://pith.science/paper/WWVJAGVI","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2407.13928&json=true","fetch_graph":"https://pith.science/api/pith-number/WWVJAGVIL4CHPE247U25J7GJIB/graph.json","fetch_events":"https://pith.science/api/pith-number/WWVJAGVIL4CHPE247U25J7GJIB/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/WWVJAGVIL4CHPE247U25J7GJIB/action/timestamp_anchor","attest_storage":"https://pith.science/pith/WWVJAGVIL4CHPE247U25J7GJIB/action/storage_attestation","attest_author":"https://pith.science/pith/WWVJAGVIL4CHPE247U25J7GJIB/action/author_attestation","sign_citation":"https://pith.science/pith/WWVJAGVIL4CHPE247U25J7GJIB/action/citation_signature","submit_replication":"https://pith.science/pith/WWVJAGVIL4CHPE247U25J7GJIB/action/replication_record"}},"created_at":"2026-07-05T08:45:55.654617+00:00","updated_at":"2026-07-05T08:45:55.654617+00:00"}