{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:EBY5VQD7LUHJ47EIMYWDJV7ESP","short_pith_number":"pith:EBY5VQD7","schema_version":"1.0","canonical_sha256":"2071dac07f5d0e9e7c88662c34d7e493c44f8b921eaeed680771d26826046131","source":{"kind":"arxiv","id":"2506.21862","version":1},"attestation_state":"computed","paper":{"title":"LLaVA-Scissor: Token Compression with Semantic Connected Components for Video LLMs","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI","cs.HC","cs.MM"],"primary_cat":"cs.CV","authors_text":"Boyuan Sun, Jiaxing Zhao, Qibin Hou, Xihan Wei","submitted_at":"2025-06-27T02:29:58Z","abstract_excerpt":"In this paper, we present LLaVA-Scissor, a training-free token compression strategy designed for video multimodal large language models. Previous methods mostly attempt to compress tokens based on attention scores, but fail to effectively capture all semantic regions and often lead to token redundancy. Differently, we propose to leverage the Semantic Connected Components (SCC) approach that assigns tokens to distinct semantic regions within the token set, ensuring comprehensive semantic coverage. The outcome is a two-step spatio-temporal token compression strategy that utilizes SCC in both spa"},"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":"2506.21862","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CV","submitted_at":"2025-06-27T02:29:58Z","cross_cats_sorted":["cs.AI","cs.HC","cs.MM"],"title_canon_sha256":"9d6d12dc89fdad67e78b8df02d21de92f6f77ee1d11e355cdf3b4a3c3cb1086c","abstract_canon_sha256":"de8788444704e4dade1b057549ae4a01bb100bcffd481790010dbd5f86e8cd6e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:28:02.913780Z","signature_b64":"7JhGnZgjbqnbf6oa6d4/ZTof7VNmLrKVVSw+85U+BIvXuLngb3PAEu1ty2qU1KCSacr+B9riW+KdL7ic7f5lBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2071dac07f5d0e9e7c88662c34d7e493c44f8b921eaeed680771d26826046131","last_reissued_at":"2026-07-05T11:28:02.913364Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:28:02.913364Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"LLaVA-Scissor: Token Compression with Semantic Connected Components for Video LLMs","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI","cs.HC","cs.MM"],"primary_cat":"cs.CV","authors_text":"Boyuan Sun, Jiaxing Zhao, Qibin Hou, Xihan Wei","submitted_at":"2025-06-27T02:29:58Z","abstract_excerpt":"In this paper, we present LLaVA-Scissor, a training-free token compression strategy designed for video multimodal large language models. Previous methods mostly attempt to compress tokens based on attention scores, but fail to effectively capture all semantic regions and often lead to token redundancy. Differently, we propose to leverage the Semantic Connected Components (SCC) approach that assigns tokens to distinct semantic regions within the token set, ensuring comprehensive semantic coverage. The outcome is a two-step spatio-temporal token compression strategy that utilizes SCC in both spa"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.21862","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/2506.21862/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":"2506.21862","created_at":"2026-07-05T11:28:02.913416+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.21862v1","created_at":"2026-07-05T11:28:02.913416+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.21862","created_at":"2026-07-05T11:28:02.913416+00:00"},{"alias_kind":"pith_short_12","alias_value":"EBY5VQD7LUHJ","created_at":"2026-07-05T11:28:02.913416+00:00"},{"alias_kind":"pith_short_16","alias_value":"EBY5VQD7LUHJ47EI","created_at":"2026-07-05T11:28:02.913416+00:00"},{"alias_kind":"pith_short_8","alias_value":"EBY5VQD7","created_at":"2026-07-05T11:28:02.913416+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.31598","citing_title":"Linear Scaling Video VLMs for Long Video Understanding","ref_index":57,"is_internal_anchor":false},{"citing_arxiv_id":"2605.18018","citing_title":"See What I Mean: Aligning Vision and Language Representations for Video Fine-grained Object Understanding","ref_index":59,"is_internal_anchor":false},{"citing_arxiv_id":"2601.01322","citing_title":"LinMU: Multimodal Understanding Made Linear","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05899","citing_title":"VisMMOE: Exploiting Visual-Expert Affinity for Efficient Visual-Language MoE Offloading","ref_index":45,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/EBY5VQD7LUHJ47EIMYWDJV7ESP","json":"https://pith.science/pith/EBY5VQD7LUHJ47EIMYWDJV7ESP.json","graph_json":"https://pith.science/api/pith-number/EBY5VQD7LUHJ47EIMYWDJV7ESP/graph.json","events_json":"https://pith.science/api/pith-number/EBY5VQD7LUHJ47EIMYWDJV7ESP/events.json","paper":"https://pith.science/paper/EBY5VQD7"},"agent_actions":{"view_html":"https://pith.science/pith/EBY5VQD7LUHJ47EIMYWDJV7ESP","download_json":"https://pith.science/pith/EBY5VQD7LUHJ47EIMYWDJV7ESP.json","view_paper":"https://pith.science/paper/EBY5VQD7","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.21862&json=true","fetch_graph":"https://pith.science/api/pith-number/EBY5VQD7LUHJ47EIMYWDJV7ESP/graph.json","fetch_events":"https://pith.science/api/pith-number/EBY5VQD7LUHJ47EIMYWDJV7ESP/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/EBY5VQD7LUHJ47EIMYWDJV7ESP/action/timestamp_anchor","attest_storage":"https://pith.science/pith/EBY5VQD7LUHJ47EIMYWDJV7ESP/action/storage_attestation","attest_author":"https://pith.science/pith/EBY5VQD7LUHJ47EIMYWDJV7ESP/action/author_attestation","sign_citation":"https://pith.science/pith/EBY5VQD7LUHJ47EIMYWDJV7ESP/action/citation_signature","submit_replication":"https://pith.science/pith/EBY5VQD7LUHJ47EIMYWDJV7ESP/action/replication_record"}},"created_at":"2026-07-05T11:28:02.913416+00:00","updated_at":"2026-07-05T11:28:02.913416+00:00"}