{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:ZPBMI3MPTWCVO2VTDPEJGSRV2I","short_pith_number":"pith:ZPBMI3MP","schema_version":"1.0","canonical_sha256":"cbc2c46d8f9d85576ab31bc8934a35d2357b54b7009a74f8c1676f7ac4afd938","source":{"kind":"arxiv","id":"2507.09184","version":2},"attestation_state":"computed","paper":{"title":"MCA-LLaVA: Manhattan Causal Attention for Reducing Hallucination in Large Vision-Language Models","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Dahan Wang, Feilong Tang, Qiyan Zhao, Sinan Fan, Xiaofeng Zhang, Xiaosong Yuan, Xuhang Chen, Xuyao Zhang, Yiheng Li, Yun Xing","submitted_at":"2025-07-12T08:09:35Z","abstract_excerpt":"Hallucinations pose a significant challenge in Large Vision Language Models (LVLMs), with misalignment between multimodal features identified as a key contributing factor. This paper reveals the negative impact of the long-term decay in Rotary Position Encoding (RoPE), used for positional modeling in LVLMs, on multimodal alignment. Concretely, under long-term decay, instruction tokens exhibit uneven perception of image tokens located at different positions within the two-dimensional space: prioritizing image tokens from the bottom-right region since in the one-dimensional sequence, these token"},"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":"2507.09184","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.CV","submitted_at":"2025-07-12T08:09:35Z","cross_cats_sorted":[],"title_canon_sha256":"af8b43d41cdb3acb4b932f3755a10ca1dfaa12b26b8cb46272845acd7c7666fd","abstract_canon_sha256":"7dc211ad2beccdd2cf0a963f1c09afa6699f9f93b04984a8b3c2cb5837c75c0a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:41:44.527145Z","signature_b64":"WjhjGFS6G1ur5cCFcMTwWXbj8QPsMJJVw2Fk0QRK6mMhgp5pIlu9iTpABJ5CqwFObWVyAWll2tvcNCQcXOxlAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"cbc2c46d8f9d85576ab31bc8934a35d2357b54b7009a74f8c1676f7ac4afd938","last_reissued_at":"2026-07-05T11:41:44.526571Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:41:44.526571Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"MCA-LLaVA: Manhattan Causal Attention for Reducing Hallucination in Large Vision-Language Models","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Dahan Wang, Feilong Tang, Qiyan Zhao, Sinan Fan, Xiaofeng Zhang, Xiaosong Yuan, Xuhang Chen, Xuyao Zhang, Yiheng Li, Yun Xing","submitted_at":"2025-07-12T08:09:35Z","abstract_excerpt":"Hallucinations pose a significant challenge in Large Vision Language Models (LVLMs), with misalignment between multimodal features identified as a key contributing factor. This paper reveals the negative impact of the long-term decay in Rotary Position Encoding (RoPE), used for positional modeling in LVLMs, on multimodal alignment. Concretely, under long-term decay, instruction tokens exhibit uneven perception of image tokens located at different positions within the two-dimensional space: prioritizing image tokens from the bottom-right region since in the one-dimensional sequence, these token"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2507.09184","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/2507.09184/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":"2507.09184","created_at":"2026-07-05T11:41:44.526636+00:00"},{"alias_kind":"arxiv_version","alias_value":"2507.09184v2","created_at":"2026-07-05T11:41:44.526636+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2507.09184","created_at":"2026-07-05T11:41:44.526636+00:00"},{"alias_kind":"pith_short_12","alias_value":"ZPBMI3MPTWCV","created_at":"2026-07-05T11:41:44.526636+00:00"},{"alias_kind":"pith_short_16","alias_value":"ZPBMI3MPTWCVO2VT","created_at":"2026-07-05T11:41:44.526636+00:00"},{"alias_kind":"pith_short_8","alias_value":"ZPBMI3MP","created_at":"2026-07-05T11:41:44.526636+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.03569","citing_title":"When Attention Collapses: Stage-Aware Visual Token Pruning from Structure to Semantics","ref_index":30,"is_internal_anchor":false},{"citing_arxiv_id":"2605.02035","citing_title":"VIDA: A dataset for Visually Dependent Ambiguity in Multimodal Machine Translation","ref_index":51,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ZPBMI3MPTWCVO2VTDPEJGSRV2I","json":"https://pith.science/pith/ZPBMI3MPTWCVO2VTDPEJGSRV2I.json","graph_json":"https://pith.science/api/pith-number/ZPBMI3MPTWCVO2VTDPEJGSRV2I/graph.json","events_json":"https://pith.science/api/pith-number/ZPBMI3MPTWCVO2VTDPEJGSRV2I/events.json","paper":"https://pith.science/paper/ZPBMI3MP"},"agent_actions":{"view_html":"https://pith.science/pith/ZPBMI3MPTWCVO2VTDPEJGSRV2I","download_json":"https://pith.science/pith/ZPBMI3MPTWCVO2VTDPEJGSRV2I.json","view_paper":"https://pith.science/paper/ZPBMI3MP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2507.09184&json=true","fetch_graph":"https://pith.science/api/pith-number/ZPBMI3MPTWCVO2VTDPEJGSRV2I/graph.json","fetch_events":"https://pith.science/api/pith-number/ZPBMI3MPTWCVO2VTDPEJGSRV2I/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ZPBMI3MPTWCVO2VTDPEJGSRV2I/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ZPBMI3MPTWCVO2VTDPEJGSRV2I/action/storage_attestation","attest_author":"https://pith.science/pith/ZPBMI3MPTWCVO2VTDPEJGSRV2I/action/author_attestation","sign_citation":"https://pith.science/pith/ZPBMI3MPTWCVO2VTDPEJGSRV2I/action/citation_signature","submit_replication":"https://pith.science/pith/ZPBMI3MPTWCVO2VTDPEJGSRV2I/action/replication_record"}},"created_at":"2026-07-05T11:41:44.526636+00:00","updated_at":"2026-07-05T11:41:44.526636+00:00"}