{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:PXGUKE732KBZJ346GQZFSIM46T","short_pith_number":"pith:PXGUKE73","schema_version":"1.0","canonical_sha256":"7dcd4513fbd28394ef9e343259219cf4cf11e2bf4e0c010103a2d36707005ccf","source":{"kind":"arxiv","id":"2506.00993","version":1},"attestation_state":"computed","paper":{"title":"FlexSelect: Flexible Token Selection for Efficient Long Video Understanding","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Fengyun Rao, Linchao Zhu, Tianyi Wang, Yi Yang, Yu Lu, Yunzhu Zhang","submitted_at":"2025-06-01T12:49:39Z","abstract_excerpt":"Long-form video understanding poses a significant challenge for video large language models (VideoLLMs) due to prohibitively high computational and memory demands. In this paper, we propose FlexSelect, a flexible and efficient token selection strategy for processing long videos. FlexSelect identifies and retains the most semantically relevant content by leveraging cross-modal attention patterns from a reference transformer layer. It comprises two key components: (1) a training-free token ranking pipeline that leverages faithful cross-modal attention weights to estimate each video token's impor"},"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":"2506.00993","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2025-06-01T12:49:39Z","cross_cats_sorted":[],"title_canon_sha256":"4646cb638e85ed08c987905925385b5ce93dd4c874d428958f8e7b863856602c","abstract_canon_sha256":"e22ebf91c80b11910319f691169285051a101d97c82ea5239afa115e950d45a2"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:13:44.168178Z","signature_b64":"b7o/4veXX4hLx2/lFgV6Je8efsWa2CoWfxCVj3hhFeIQYOImGMWgg8N5PCRWy5qa7ip1+yA7ukMiTa83LVvWAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7dcd4513fbd28394ef9e343259219cf4cf11e2bf4e0c010103a2d36707005ccf","last_reissued_at":"2026-07-05T11:13:44.167752Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:13:44.167752Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"FlexSelect: Flexible Token Selection for Efficient Long Video Understanding","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Fengyun Rao, Linchao Zhu, Tianyi Wang, Yi Yang, Yu Lu, Yunzhu Zhang","submitted_at":"2025-06-01T12:49:39Z","abstract_excerpt":"Long-form video understanding poses a significant challenge for video large language models (VideoLLMs) due to prohibitively high computational and memory demands. In this paper, we propose FlexSelect, a flexible and efficient token selection strategy for processing long videos. FlexSelect identifies and retains the most semantically relevant content by leveraging cross-modal attention patterns from a reference transformer layer. It comprises two key components: (1) a training-free token ranking pipeline that leverages faithful cross-modal attention weights to estimate each video token's impor"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.00993","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/2506.00993/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":"2506.00993","created_at":"2026-07-05T11:13:44.167810+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.00993v1","created_at":"2026-07-05T11:13:44.167810+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.00993","created_at":"2026-07-05T11:13:44.167810+00:00"},{"alias_kind":"pith_short_12","alias_value":"PXGUKE732KBZ","created_at":"2026-07-05T11:13:44.167810+00:00"},{"alias_kind":"pith_short_16","alias_value":"PXGUKE732KBZJ346","created_at":"2026-07-05T11:13:44.167810+00:00"},{"alias_kind":"pith_short_8","alias_value":"PXGUKE73","created_at":"2026-07-05T11:13:44.167810+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.27600","citing_title":"Purifying Multimodal Retrieval: Fragment-Level Evidence Selection for RAG","ref_index":56,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/PXGUKE732KBZJ346GQZFSIM46T","json":"https://pith.science/pith/PXGUKE732KBZJ346GQZFSIM46T.json","graph_json":"https://pith.science/api/pith-number/PXGUKE732KBZJ346GQZFSIM46T/graph.json","events_json":"https://pith.science/api/pith-number/PXGUKE732KBZJ346GQZFSIM46T/events.json","paper":"https://pith.science/paper/PXGUKE73"},"agent_actions":{"view_html":"https://pith.science/pith/PXGUKE732KBZJ346GQZFSIM46T","download_json":"https://pith.science/pith/PXGUKE732KBZJ346GQZFSIM46T.json","view_paper":"https://pith.science/paper/PXGUKE73","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.00993&json=true","fetch_graph":"https://pith.science/api/pith-number/PXGUKE732KBZJ346GQZFSIM46T/graph.json","fetch_events":"https://pith.science/api/pith-number/PXGUKE732KBZJ346GQZFSIM46T/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PXGUKE732KBZJ346GQZFSIM46T/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PXGUKE732KBZJ346GQZFSIM46T/action/storage_attestation","attest_author":"https://pith.science/pith/PXGUKE732KBZJ346GQZFSIM46T/action/author_attestation","sign_citation":"https://pith.science/pith/PXGUKE732KBZJ346GQZFSIM46T/action/citation_signature","submit_replication":"https://pith.science/pith/PXGUKE732KBZJ346GQZFSIM46T/action/replication_record"}},"created_at":"2026-07-05T11:13:44.167810+00:00","updated_at":"2026-07-05T11:13:44.167810+00:00"}