{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:YDLIVOT5SCYZRF4KLMTGK2FZFC","short_pith_number":"pith:YDLIVOT5","schema_version":"1.0","canonical_sha256":"c0d68aba7d90b198978a5b266568b928b98dc0c021c1e16af3df7eb0cf9e3e3a","source":{"kind":"arxiv","id":"2404.10501","version":2},"attestation_state":"computed","paper":{"title":"Self-Supervised Visual Preference Alignment","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.LG"],"primary_cat":"cs.CV","authors_text":"Ke Zhu, Liang Zhao, Xiangyu Zhang, Zheng Ge","submitted_at":"2024-04-16T12:19:54Z","abstract_excerpt":"This paper makes the first attempt towards unsupervised preference alignment in Vision-Language Models (VLMs). We generate chosen and rejected responses with regard to the original and augmented image pairs, and conduct preference alignment with direct preference optimization. It is based on a core idea: properly designed augmentation to the image input will induce VLM to generate false but hard negative responses, which helps the model to learn from and produce more robust and powerful answers. The whole pipeline no longer hinges on supervision from GPT-4 or human involvement during alignment"},"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":"2404.10501","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2024-04-16T12:19:54Z","cross_cats_sorted":["cs.AI","cs.CL","cs.LG"],"title_canon_sha256":"43162daf8f3221fa45149bdd8a0776fc5220f3559938c51094c9a6a8d2d54e04","abstract_canon_sha256":"11570ec4dccc25edf5be7d2f36505d4089cac20253a592021fa3a3fbb24ca0e1"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:57:37.606694Z","signature_b64":"LlHczMJOA7YwwkPADSCTCno1jcL3ndhjYxo0KHTIwTxcF0LpCTdVfLrAHDitifBADyBs4mb6vahZ8Ijzoq2BBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c0d68aba7d90b198978a5b266568b928b98dc0c021c1e16af3df7eb0cf9e3e3a","last_reissued_at":"2026-07-05T08:57:37.606145Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:57:37.606145Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Self-Supervised Visual Preference Alignment","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.LG"],"primary_cat":"cs.CV","authors_text":"Ke Zhu, Liang Zhao, Xiangyu Zhang, Zheng Ge","submitted_at":"2024-04-16T12:19:54Z","abstract_excerpt":"This paper makes the first attempt towards unsupervised preference alignment in Vision-Language Models (VLMs). We generate chosen and rejected responses with regard to the original and augmented image pairs, and conduct preference alignment with direct preference optimization. It is based on a core idea: properly designed augmentation to the image input will induce VLM to generate false but hard negative responses, which helps the model to learn from and produce more robust and powerful answers. The whole pipeline no longer hinges on supervision from GPT-4 or human involvement during alignment"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2404.10501","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/2404.10501/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":"2404.10501","created_at":"2026-07-05T08:57:37.606221+00:00"},{"alias_kind":"arxiv_version","alias_value":"2404.10501v2","created_at":"2026-07-05T08:57:37.606221+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2404.10501","created_at":"2026-07-05T08:57:37.606221+00:00"},{"alias_kind":"pith_short_12","alias_value":"YDLIVOT5SCYZ","created_at":"2026-07-05T08:57:37.606221+00:00"},{"alias_kind":"pith_short_16","alias_value":"YDLIVOT5SCYZRF4K","created_at":"2026-07-05T08:57:37.606221+00:00"},{"alias_kind":"pith_short_8","alias_value":"YDLIVOT5","created_at":"2026-07-05T08:57:37.606221+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.18512","citing_title":"S2H-DPO: Hardness-Aware Preference Optimization for Vision-Language Models","ref_index":65,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/YDLIVOT5SCYZRF4KLMTGK2FZFC","json":"https://pith.science/pith/YDLIVOT5SCYZRF4KLMTGK2FZFC.json","graph_json":"https://pith.science/api/pith-number/YDLIVOT5SCYZRF4KLMTGK2FZFC/graph.json","events_json":"https://pith.science/api/pith-number/YDLIVOT5SCYZRF4KLMTGK2FZFC/events.json","paper":"https://pith.science/paper/YDLIVOT5"},"agent_actions":{"view_html":"https://pith.science/pith/YDLIVOT5SCYZRF4KLMTGK2FZFC","download_json":"https://pith.science/pith/YDLIVOT5SCYZRF4KLMTGK2FZFC.json","view_paper":"https://pith.science/paper/YDLIVOT5","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2404.10501&json=true","fetch_graph":"https://pith.science/api/pith-number/YDLIVOT5SCYZRF4KLMTGK2FZFC/graph.json","fetch_events":"https://pith.science/api/pith-number/YDLIVOT5SCYZRF4KLMTGK2FZFC/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/YDLIVOT5SCYZRF4KLMTGK2FZFC/action/timestamp_anchor","attest_storage":"https://pith.science/pith/YDLIVOT5SCYZRF4KLMTGK2FZFC/action/storage_attestation","attest_author":"https://pith.science/pith/YDLIVOT5SCYZRF4KLMTGK2FZFC/action/author_attestation","sign_citation":"https://pith.science/pith/YDLIVOT5SCYZRF4KLMTGK2FZFC/action/citation_signature","submit_replication":"https://pith.science/pith/YDLIVOT5SCYZRF4KLMTGK2FZFC/action/replication_record"}},"created_at":"2026-07-05T08:57:37.606221+00:00","updated_at":"2026-07-05T08:57:37.606221+00:00"}