{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:NJVJJDOV5EPCDZXOBD3HPG3GIR","short_pith_number":"pith:NJVJJDOV","schema_version":"1.0","canonical_sha256":"6a6a948dd5e91e21e6ee08f6779b66447bd8f9cb7a04d4e7e49d84cade664607","source":{"kind":"arxiv","id":"2605.31312","version":1},"attestation_state":"computed","paper":{"title":"Learning from Fine-Grained Visual Discrepancies: Mitigating Multimodal Hallucinations via In-Context Visual Contrastive Optimization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.CV","authors_text":"Chen Chen, Haolin Deng, Haonan Lu, Xin Zou, Xuming Hu, Zhiwei Jin","submitted_at":"2026-05-29T13:44:56Z","abstract_excerpt":"Multimodal hallucination remains a persistent challenge for Vision-Language Models (VLMs). Standard textual Direct Preference Optimization (DPO) often fails to mitigate it due to a lack of explicit visual supervision. While existing works introduce visual preference DPO by contrasting original images against negative ones, they suffer from a theoretically inconsistent objective caused by partition function mismatches and rely on coarse-grained negatives that could enable shortcut learning. In this work, we propose In-Context Visual Contrastive Optimization (IC-VCO). By placing contrastive imag"},"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":"2605.31312","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-05-29T13:44:56Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"2e876e60467288a297a5abf8412eddf1442c165e4d863b812e52882bb6333556","abstract_canon_sha256":"0369cff2b46dfc98f73941001b6b470924691358df0aba2931e2bdb02bcc9de3"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-01T01:04:10.331591Z","signature_b64":"TSx/EYjrpDr3SzmdPGvF3JCRJlIcfI6+Ju4mBMH9+AEmh3c5QuxvlsuBbu0r25D/56L1TW7PiMNq2zUGPTt7Bg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6a6a948dd5e91e21e6ee08f6779b66447bd8f9cb7a04d4e7e49d84cade664607","last_reissued_at":"2026-06-01T01:04:10.330957Z","signature_status":"signed_v1","first_computed_at":"2026-06-01T01:04:10.330957Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Learning from Fine-Grained Visual Discrepancies: Mitigating Multimodal Hallucinations via In-Context Visual Contrastive Optimization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.CV","authors_text":"Chen Chen, Haolin Deng, Haonan Lu, Xin Zou, Xuming Hu, Zhiwei Jin","submitted_at":"2026-05-29T13:44:56Z","abstract_excerpt":"Multimodal hallucination remains a persistent challenge for Vision-Language Models (VLMs). Standard textual Direct Preference Optimization (DPO) often fails to mitigate it due to a lack of explicit visual supervision. While existing works introduce visual preference DPO by contrasting original images against negative ones, they suffer from a theoretically inconsistent objective caused by partition function mismatches and rely on coarse-grained negatives that could enable shortcut learning. In this work, we propose In-Context Visual Contrastive Optimization (IC-VCO). By placing contrastive imag"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.31312","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/2605.31312/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":"2605.31312","created_at":"2026-06-01T01:04:10.331057+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.31312v1","created_at":"2026-06-01T01:04:10.331057+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.31312","created_at":"2026-06-01T01:04:10.331057+00:00"},{"alias_kind":"pith_short_12","alias_value":"NJVJJDOV5EPC","created_at":"2026-06-01T01:04:10.331057+00:00"},{"alias_kind":"pith_short_16","alias_value":"NJVJJDOV5EPCDZXO","created_at":"2026-06-01T01:04:10.331057+00:00"},{"alias_kind":"pith_short_8","alias_value":"NJVJJDOV","created_at":"2026-06-01T01:04:10.331057+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/NJVJJDOV5EPCDZXOBD3HPG3GIR","json":"https://pith.science/pith/NJVJJDOV5EPCDZXOBD3HPG3GIR.json","graph_json":"https://pith.science/api/pith-number/NJVJJDOV5EPCDZXOBD3HPG3GIR/graph.json","events_json":"https://pith.science/api/pith-number/NJVJJDOV5EPCDZXOBD3HPG3GIR/events.json","paper":"https://pith.science/paper/NJVJJDOV"},"agent_actions":{"view_html":"https://pith.science/pith/NJVJJDOV5EPCDZXOBD3HPG3GIR","download_json":"https://pith.science/pith/NJVJJDOV5EPCDZXOBD3HPG3GIR.json","view_paper":"https://pith.science/paper/NJVJJDOV","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.31312&json=true","fetch_graph":"https://pith.science/api/pith-number/NJVJJDOV5EPCDZXOBD3HPG3GIR/graph.json","fetch_events":"https://pith.science/api/pith-number/NJVJJDOV5EPCDZXOBD3HPG3GIR/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/NJVJJDOV5EPCDZXOBD3HPG3GIR/action/timestamp_anchor","attest_storage":"https://pith.science/pith/NJVJJDOV5EPCDZXOBD3HPG3GIR/action/storage_attestation","attest_author":"https://pith.science/pith/NJVJJDOV5EPCDZXOBD3HPG3GIR/action/author_attestation","sign_citation":"https://pith.science/pith/NJVJJDOV5EPCDZXOBD3HPG3GIR/action/citation_signature","submit_replication":"https://pith.science/pith/NJVJJDOV5EPCDZXOBD3HPG3GIR/action/replication_record"}},"created_at":"2026-06-01T01:04:10.331057+00:00","updated_at":"2026-06-01T01:04:10.331057+00:00"}