{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:6ZMPGMXGFSS5UIOZ32NDX56TGL","short_pith_number":"pith:6ZMPGMXG","schema_version":"1.0","canonical_sha256":"f658f332e62ca5da21d9de9a3bf7d332fc3e73aad1e00ab3f151a86c8b6f377d","source":{"kind":"arxiv","id":"2503.17928","version":1},"attestation_state":"computed","paper":{"title":"Debiasing Multimodal Large Language Models via Noise-Aware Preference Optimization","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.CV","authors_text":"Dawei Yin, Duohe Ma, Hengzhu Tang, Jiawei Sheng, Tingwen Liu, Yiming Ren, Zefeng Zhang, Zhenyang Li, Zhenyu Zhang","submitted_at":"2025-03-23T04:00:11Z","abstract_excerpt":"Multimodal Large Language Models excel in various tasks, yet often struggle with modality bias, where the model tends to rely heavily on a single modality and overlook critical information in other modalities, which leads to incorrect focus and generating irrelevant responses. In this paper, we propose using the paradigm of preference optimization to solve the modality bias problem, including RLAIFVBias, a debiased preference optimization dataset, and a Noise Aware Preference Optimization algorithm. Specifically, we first construct the dataset by introducing perturbations to reduce the informa"},"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":"2503.17928","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2025-03-23T04:00:11Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"c5a5261343fa923b6457035caaabc1911ca87b379ba084c619ba2e7b31025c4e","abstract_canon_sha256":"436ee76437120803142dfb1b4ef260489aa5b1684165f6c08675215fc53d122d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:37:58.847237Z","signature_b64":"ft8t0MpXyQ9BtfZBXyRtuJzD+59J4T42l+apSwGvyTMB/buGZ8LX4RZHXQA9cVBDSZetPbQMvGhQUpUxFZO0AA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f658f332e62ca5da21d9de9a3bf7d332fc3e73aad1e00ab3f151a86c8b6f377d","last_reissued_at":"2026-07-05T10:37:58.846528Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:37:58.846528Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Debiasing Multimodal Large Language Models via Noise-Aware Preference Optimization","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.CV","authors_text":"Dawei Yin, Duohe Ma, Hengzhu Tang, Jiawei Sheng, Tingwen Liu, Yiming Ren, Zefeng Zhang, Zhenyang Li, Zhenyu Zhang","submitted_at":"2025-03-23T04:00:11Z","abstract_excerpt":"Multimodal Large Language Models excel in various tasks, yet often struggle with modality bias, where the model tends to rely heavily on a single modality and overlook critical information in other modalities, which leads to incorrect focus and generating irrelevant responses. In this paper, we propose using the paradigm of preference optimization to solve the modality bias problem, including RLAIFVBias, a debiased preference optimization dataset, and a Noise Aware Preference Optimization algorithm. Specifically, we first construct the dataset by introducing perturbations to reduce the informa"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.17928","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/2503.17928/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":"2503.17928","created_at":"2026-07-05T10:37:58.846616+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.17928v1","created_at":"2026-07-05T10:37:58.846616+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.17928","created_at":"2026-07-05T10:37:58.846616+00:00"},{"alias_kind":"pith_short_12","alias_value":"6ZMPGMXGFSS5","created_at":"2026-07-05T10:37:58.846616+00:00"},{"alias_kind":"pith_short_16","alias_value":"6ZMPGMXGFSS5UIOZ","created_at":"2026-07-05T10:37:58.846616+00:00"},{"alias_kind":"pith_short_8","alias_value":"6ZMPGMXG","created_at":"2026-07-05T10:37:58.846616+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.20366","citing_title":"Mitigating Hallucinations in Large Vision-Language Models without Performance Degradation","ref_index":171,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/6ZMPGMXGFSS5UIOZ32NDX56TGL","json":"https://pith.science/pith/6ZMPGMXGFSS5UIOZ32NDX56TGL.json","graph_json":"https://pith.science/api/pith-number/6ZMPGMXGFSS5UIOZ32NDX56TGL/graph.json","events_json":"https://pith.science/api/pith-number/6ZMPGMXGFSS5UIOZ32NDX56TGL/events.json","paper":"https://pith.science/paper/6ZMPGMXG"},"agent_actions":{"view_html":"https://pith.science/pith/6ZMPGMXGFSS5UIOZ32NDX56TGL","download_json":"https://pith.science/pith/6ZMPGMXGFSS5UIOZ32NDX56TGL.json","view_paper":"https://pith.science/paper/6ZMPGMXG","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.17928&json=true","fetch_graph":"https://pith.science/api/pith-number/6ZMPGMXGFSS5UIOZ32NDX56TGL/graph.json","fetch_events":"https://pith.science/api/pith-number/6ZMPGMXGFSS5UIOZ32NDX56TGL/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/6ZMPGMXGFSS5UIOZ32NDX56TGL/action/timestamp_anchor","attest_storage":"https://pith.science/pith/6ZMPGMXGFSS5UIOZ32NDX56TGL/action/storage_attestation","attest_author":"https://pith.science/pith/6ZMPGMXGFSS5UIOZ32NDX56TGL/action/author_attestation","sign_citation":"https://pith.science/pith/6ZMPGMXGFSS5UIOZ32NDX56TGL/action/citation_signature","submit_replication":"https://pith.science/pith/6ZMPGMXGFSS5UIOZ32NDX56TGL/action/replication_record"}},"created_at":"2026-07-05T10:37:58.846616+00:00","updated_at":"2026-07-05T10:37:58.846616+00:00"}