{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:Y5LTLPYDBRPLYUR37L6DQCIMOS","short_pith_number":"pith:Y5LTLPYD","schema_version":"1.0","canonical_sha256":"c75735bf030c5ebc523bfafc38090c74a4267991111f24264eedbdd58db0f351","source":{"kind":"arxiv","id":"2606.31054","version":1},"attestation_state":"computed","paper":{"title":"ADAPT: Attention Dynamics Alignment with Preference Tuning for Faithful MLLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.MM"],"primary_cat":"cs.CV","authors_text":"Jiajun Li, Yi Tu, Zhendong Mao, Zheren Fu, Zhixiao Zheng, Zhiyuan Yao","submitted_at":"2026-06-30T02:46:10Z","abstract_excerpt":"Multimodal Large Language Models (MLLMs) are critically hampered by hallucination, generating content inconsistent with the provided image. In this paper, we identify an internal signature of hallucination: progressive degradation of text-to-image cross-attention during generation, leading to specific failure patterns like unfocused or biased attention. Existing mitigation strategies are largely outcome-driven and do not explicitly target this failure mode. To address this problem, we propose ADAPT (Attention Dynamics Alignment with Preference Tuning), an attention-based framework that interve"},"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":"2606.31054","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-06-30T02:46:10Z","cross_cats_sorted":["cs.AI","cs.CL","cs.MM"],"title_canon_sha256":"7af8acd2440bf56b9638f93ef7aeced329232261d9cee7684c70d7d4ce28b234","abstract_canon_sha256":"fea153fc4aec48c8c76899b81a3d5cdaf43e3655c5c80b350df6c4da271a7d66"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-01T01:17:27.732083Z","signature_b64":"yvBZvCWXCUqcTlVD0exWrCq3DzOOq1vJXQty82SuI7wDdFo3DzquQRREI7Whnf8MPboWPWc74GVfkF7H5fK9BA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c75735bf030c5ebc523bfafc38090c74a4267991111f24264eedbdd58db0f351","last_reissued_at":"2026-07-01T01:17:27.731653Z","signature_status":"signed_v1","first_computed_at":"2026-07-01T01:17:27.731653Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"ADAPT: Attention Dynamics Alignment with Preference Tuning for Faithful MLLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.MM"],"primary_cat":"cs.CV","authors_text":"Jiajun Li, Yi Tu, Zhendong Mao, Zheren Fu, Zhixiao Zheng, Zhiyuan Yao","submitted_at":"2026-06-30T02:46:10Z","abstract_excerpt":"Multimodal Large Language Models (MLLMs) are critically hampered by hallucination, generating content inconsistent with the provided image. In this paper, we identify an internal signature of hallucination: progressive degradation of text-to-image cross-attention during generation, leading to specific failure patterns like unfocused or biased attention. Existing mitigation strategies are largely outcome-driven and do not explicitly target this failure mode. To address this problem, we propose ADAPT (Attention Dynamics Alignment with Preference Tuning), an attention-based framework that interve"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.31054","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/2606.31054/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":"2606.31054","created_at":"2026-07-01T01:17:27.731716+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.31054v1","created_at":"2026-07-01T01:17:27.731716+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.31054","created_at":"2026-07-01T01:17:27.731716+00:00"},{"alias_kind":"pith_short_12","alias_value":"Y5LTLPYDBRPL","created_at":"2026-07-01T01:17:27.731716+00:00"},{"alias_kind":"pith_short_16","alias_value":"Y5LTLPYDBRPLYUR3","created_at":"2026-07-01T01:17:27.731716+00:00"},{"alias_kind":"pith_short_8","alias_value":"Y5LTLPYD","created_at":"2026-07-01T01:17:27.731716+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/Y5LTLPYDBRPLYUR37L6DQCIMOS","json":"https://pith.science/pith/Y5LTLPYDBRPLYUR37L6DQCIMOS.json","graph_json":"https://pith.science/api/pith-number/Y5LTLPYDBRPLYUR37L6DQCIMOS/graph.json","events_json":"https://pith.science/api/pith-number/Y5LTLPYDBRPLYUR37L6DQCIMOS/events.json","paper":"https://pith.science/paper/Y5LTLPYD"},"agent_actions":{"view_html":"https://pith.science/pith/Y5LTLPYDBRPLYUR37L6DQCIMOS","download_json":"https://pith.science/pith/Y5LTLPYDBRPLYUR37L6DQCIMOS.json","view_paper":"https://pith.science/paper/Y5LTLPYD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.31054&json=true","fetch_graph":"https://pith.science/api/pith-number/Y5LTLPYDBRPLYUR37L6DQCIMOS/graph.json","fetch_events":"https://pith.science/api/pith-number/Y5LTLPYDBRPLYUR37L6DQCIMOS/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/Y5LTLPYDBRPLYUR37L6DQCIMOS/action/timestamp_anchor","attest_storage":"https://pith.science/pith/Y5LTLPYDBRPLYUR37L6DQCIMOS/action/storage_attestation","attest_author":"https://pith.science/pith/Y5LTLPYDBRPLYUR37L6DQCIMOS/action/author_attestation","sign_citation":"https://pith.science/pith/Y5LTLPYDBRPLYUR37L6DQCIMOS/action/citation_signature","submit_replication":"https://pith.science/pith/Y5LTLPYDBRPLYUR37L6DQCIMOS/action/replication_record"}},"created_at":"2026-07-01T01:17:27.731716+00:00","updated_at":"2026-07-01T01:17:27.731716+00:00"}