{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:4RYX6ACMVGDFJUWBNGOI7CDC7E","short_pith_number":"pith:4RYX6ACM","schema_version":"1.0","canonical_sha256":"e4717f004ca98654d2c1699c8f8862f90419b96e36a5acc34b6686c617cf29e7","source":{"kind":"arxiv","id":"2502.02406","version":3},"attestation_state":"computed","paper":{"title":"LV-XAttn: Distributed Cross-Attention for Long Visual Inputs in Multimodal Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.DC","cs.LG"],"primary_cat":"cs.CV","authors_text":"Shivaram Venkataraman, Tzu-Tao Chang","submitted_at":"2025-02-04T15:24:16Z","abstract_excerpt":"Cross-attention is commonly adopted in multimodal large language models (MLLMs) for integrating visual information into the language backbone. However, in applications with large visual inputs, such as video understanding, processing a large number of visual tokens in cross-attention layers leads to high memory demands and often necessitates distributed computation across multiple GPUs. Existing distributed attention mechanisms face significant communication overheads, making cross-attention layers a critical bottleneck for efficient training and inference of MLLMs. To address this, we propose"},"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":"2502.02406","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2025-02-04T15:24:16Z","cross_cats_sorted":["cs.AI","cs.DC","cs.LG"],"title_canon_sha256":"afa6cdc43485e6aaebb2ca5c6da194aff679b8486c59ff9adb301b7615ec5f9f","abstract_canon_sha256":"d696a216515805a0b5ae8de82faeb677b3d68183ccede8e43919396326f7b706"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:10:51.246395Z","signature_b64":"HWTBs2D6yOfFsHnT6BG3o0HPeisVc755gMhuGboX8Tb2ffVLqOIUU19LJlLJ7S/CM4dNXjSVyfSzYUsuE4rrCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e4717f004ca98654d2c1699c8f8862f90419b96e36a5acc34b6686c617cf29e7","last_reissued_at":"2026-07-05T11:10:51.245921Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:10:51.245921Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"LV-XAttn: Distributed Cross-Attention for Long Visual Inputs in Multimodal Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.DC","cs.LG"],"primary_cat":"cs.CV","authors_text":"Shivaram Venkataraman, Tzu-Tao Chang","submitted_at":"2025-02-04T15:24:16Z","abstract_excerpt":"Cross-attention is commonly adopted in multimodal large language models (MLLMs) for integrating visual information into the language backbone. However, in applications with large visual inputs, such as video understanding, processing a large number of visual tokens in cross-attention layers leads to high memory demands and often necessitates distributed computation across multiple GPUs. Existing distributed attention mechanisms face significant communication overheads, making cross-attention layers a critical bottleneck for efficient training and inference of MLLMs. To address this, we propose"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.02406","kind":"arxiv","version":3},"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/2502.02406/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":"2502.02406","created_at":"2026-07-05T11:10:51.245974+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.02406v3","created_at":"2026-07-05T11:10:51.245974+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.02406","created_at":"2026-07-05T11:10:51.245974+00:00"},{"alias_kind":"pith_short_12","alias_value":"4RYX6ACMVGDF","created_at":"2026-07-05T11:10:51.245974+00:00"},{"alias_kind":"pith_short_16","alias_value":"4RYX6ACMVGDFJUWB","created_at":"2026-07-05T11:10:51.245974+00:00"},{"alias_kind":"pith_short_8","alias_value":"4RYX6ACM","created_at":"2026-07-05T11:10:51.245974+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/4RYX6ACMVGDFJUWBNGOI7CDC7E","json":"https://pith.science/pith/4RYX6ACMVGDFJUWBNGOI7CDC7E.json","graph_json":"https://pith.science/api/pith-number/4RYX6ACMVGDFJUWBNGOI7CDC7E/graph.json","events_json":"https://pith.science/api/pith-number/4RYX6ACMVGDFJUWBNGOI7CDC7E/events.json","paper":"https://pith.science/paper/4RYX6ACM"},"agent_actions":{"view_html":"https://pith.science/pith/4RYX6ACMVGDFJUWBNGOI7CDC7E","download_json":"https://pith.science/pith/4RYX6ACMVGDFJUWBNGOI7CDC7E.json","view_paper":"https://pith.science/paper/4RYX6ACM","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.02406&json=true","fetch_graph":"https://pith.science/api/pith-number/4RYX6ACMVGDFJUWBNGOI7CDC7E/graph.json","fetch_events":"https://pith.science/api/pith-number/4RYX6ACMVGDFJUWBNGOI7CDC7E/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/4RYX6ACMVGDFJUWBNGOI7CDC7E/action/timestamp_anchor","attest_storage":"https://pith.science/pith/4RYX6ACMVGDFJUWBNGOI7CDC7E/action/storage_attestation","attest_author":"https://pith.science/pith/4RYX6ACMVGDFJUWBNGOI7CDC7E/action/author_attestation","sign_citation":"https://pith.science/pith/4RYX6ACMVGDFJUWBNGOI7CDC7E/action/citation_signature","submit_replication":"https://pith.science/pith/4RYX6ACMVGDFJUWBNGOI7CDC7E/action/replication_record"}},"created_at":"2026-07-05T11:10:51.245974+00:00","updated_at":"2026-07-05T11:10:51.245974+00:00"}