{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:PIBQGAZQMQSLNWP2V2K7PN35CV","short_pith_number":"pith:PIBQGAZQ","schema_version":"1.0","canonical_sha256":"7a030303306424b6d9faae95f7b77d156ff74890bc34822d16557cfea18a2c86","source":{"kind":"arxiv","id":"2606.08511","version":1},"attestation_state":"computed","paper":{"title":"Look Less, Reason More: Block-wise Attention Skipping for Efficient Multimodal LLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Jiayi Ji, Jie Ma, Rongrong Ji, Xiaoshuai Sun, Zhike Qiu","submitted_at":"2026-06-07T08:32:13Z","abstract_excerpt":"Multimodal Large Language Models (MLLMs) face a significant inference bottleneck due to the quadratic computational cost of self-attention over long visual token sequences. However, we identify a critical inefficiency in current architectures: Visual Attention Saturation. Our analysis reveals that visual tokens rapidly establish their spatial structure and intra-modal relationships in early layers, rendering visual-to-visual self-attention in deeper layers computationally redundant. Conversely, Feed-Forward Networks (FFNs) in these layers remain essential for projecting visual features into th"},"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":"2606.08511","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-06-07T08:32:13Z","cross_cats_sorted":[],"title_canon_sha256":"4159dc75cb61dd3fd7a52517c6c4db2581a018f7521fc157f7a56889db798927","abstract_canon_sha256":"9db3817a646a9488a2536ccd3dd2796c2555279d055f460a2ebdc1754f09fc5a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-09T01:05:38.678976Z","signature_b64":"g3djTQ/n/6vwKO9r941kfDeOKwVBWG+Cs4jNr3T/3H5ylUWGdzpirRCQdNo5gkDhBYbTu29+DIfOzyZ6JnGcBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7a030303306424b6d9faae95f7b77d156ff74890bc34822d16557cfea18a2c86","last_reissued_at":"2026-06-09T01:05:38.678574Z","signature_status":"signed_v1","first_computed_at":"2026-06-09T01:05:38.678574Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Look Less, Reason More: Block-wise Attention Skipping for Efficient Multimodal LLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Jiayi Ji, Jie Ma, Rongrong Ji, Xiaoshuai Sun, Zhike Qiu","submitted_at":"2026-06-07T08:32:13Z","abstract_excerpt":"Multimodal Large Language Models (MLLMs) face a significant inference bottleneck due to the quadratic computational cost of self-attention over long visual token sequences. However, we identify a critical inefficiency in current architectures: Visual Attention Saturation. Our analysis reveals that visual tokens rapidly establish their spatial structure and intra-modal relationships in early layers, rendering visual-to-visual self-attention in deeper layers computationally redundant. Conversely, Feed-Forward Networks (FFNs) in these layers remain essential for projecting visual features into th"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.08511","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/2606.08511/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":"2606.08511","created_at":"2026-06-09T01:05:38.678634+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.08511v1","created_at":"2026-06-09T01:05:38.678634+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.08511","created_at":"2026-06-09T01:05:38.678634+00:00"},{"alias_kind":"pith_short_12","alias_value":"PIBQGAZQMQSL","created_at":"2026-06-09T01:05:38.678634+00:00"},{"alias_kind":"pith_short_16","alias_value":"PIBQGAZQMQSLNWP2","created_at":"2026-06-09T01:05:38.678634+00:00"},{"alias_kind":"pith_short_8","alias_value":"PIBQGAZQ","created_at":"2026-06-09T01:05:38.678634+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/PIBQGAZQMQSLNWP2V2K7PN35CV","json":"https://pith.science/pith/PIBQGAZQMQSLNWP2V2K7PN35CV.json","graph_json":"https://pith.science/api/pith-number/PIBQGAZQMQSLNWP2V2K7PN35CV/graph.json","events_json":"https://pith.science/api/pith-number/PIBQGAZQMQSLNWP2V2K7PN35CV/events.json","paper":"https://pith.science/paper/PIBQGAZQ"},"agent_actions":{"view_html":"https://pith.science/pith/PIBQGAZQMQSLNWP2V2K7PN35CV","download_json":"https://pith.science/pith/PIBQGAZQMQSLNWP2V2K7PN35CV.json","view_paper":"https://pith.science/paper/PIBQGAZQ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.08511&json=true","fetch_graph":"https://pith.science/api/pith-number/PIBQGAZQMQSLNWP2V2K7PN35CV/graph.json","fetch_events":"https://pith.science/api/pith-number/PIBQGAZQMQSLNWP2V2K7PN35CV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PIBQGAZQMQSLNWP2V2K7PN35CV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PIBQGAZQMQSLNWP2V2K7PN35CV/action/storage_attestation","attest_author":"https://pith.science/pith/PIBQGAZQMQSLNWP2V2K7PN35CV/action/author_attestation","sign_citation":"https://pith.science/pith/PIBQGAZQMQSLNWP2V2K7PN35CV/action/citation_signature","submit_replication":"https://pith.science/pith/PIBQGAZQMQSLNWP2V2K7PN35CV/action/replication_record"}},"created_at":"2026-06-09T01:05:38.678634+00:00","updated_at":"2026-06-09T01:05:38.678634+00:00"}