{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:WQPC24LQMSJ3INWP4U6IC4ADHO","short_pith_number":"pith:WQPC24LQ","schema_version":"1.0","canonical_sha256":"b41e2d71706493b436cfe53c8170033b912a6ac95254d9714affa1de05b23e5a","source":{"kind":"arxiv","id":"2605.20035","version":1},"attestation_state":"computed","paper":{"title":"Stage-adaptive Token Selection for Efficient Omni-modal LLMs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Fengyun Rao, Jie Yang, Jing Lyu, Ruixiang Zhao, Tianyi Wang, Xirong Li, Zijie Xin","submitted_at":"2026-05-19T15:55:16Z","abstract_excerpt":"Omni-modal large language models (om-LLMs) achieve unified audio-visual understanding by encoding video and audio into temporally aligned token sequences interleaved at the window level. However, processing these dense non-textual tokens throughout the LLM incurs substantial computational overhead. Although training-free token selection can reduce this cost, existing methods either focus on visual-only inputs or prune om-LLM tokens only before the LLM with fixed per-modality ratios, failing to capture how cross-modal token importance evolves across layers. To address this limitation, we first "},"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":"2605.20035","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-05-19T15:55:16Z","cross_cats_sorted":[],"title_canon_sha256":"65ad8c0ba812f6fbabd6a76b6621e2298a6925a70de6a7a1b9345ed6d38b2ce1","abstract_canon_sha256":"d952448f510c17ba7b2ea7e536c8e7510383dcfa4856106396dc4482a28544c4"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T02:05:59.716889Z","signature_b64":"u80niZX1zelQJVh75PuPFi1ZGTz7KKXrzmaZOsnQ/fLXvaQQO6n0HHPj+Y129e8v8jA3uDuV+AydOYPJbnFyAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b41e2d71706493b436cfe53c8170033b912a6ac95254d9714affa1de05b23e5a","last_reissued_at":"2026-05-20T02:05:59.716446Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T02:05:59.716446Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Stage-adaptive Token Selection for Efficient Omni-modal LLMs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Fengyun Rao, Jie Yang, Jing Lyu, Ruixiang Zhao, Tianyi Wang, Xirong Li, Zijie Xin","submitted_at":"2026-05-19T15:55:16Z","abstract_excerpt":"Omni-modal large language models (om-LLMs) achieve unified audio-visual understanding by encoding video and audio into temporally aligned token sequences interleaved at the window level. However, processing these dense non-textual tokens throughout the LLM incurs substantial computational overhead. Although training-free token selection can reduce this cost, existing methods either focus on visual-only inputs or prune om-LLM tokens only before the LLM with fixed per-modality ratios, failing to capture how cross-modal token importance evolves across layers. To address this limitation, we first "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.20035","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/2605.20035/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":"2605.20035","created_at":"2026-05-20T02:05:59.716514+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.20035v1","created_at":"2026-05-20T02:05:59.716514+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.20035","created_at":"2026-05-20T02:05:59.716514+00:00"},{"alias_kind":"pith_short_12","alias_value":"WQPC24LQMSJ3","created_at":"2026-05-20T02:05:59.716514+00:00"},{"alias_kind":"pith_short_16","alias_value":"WQPC24LQMSJ3INWP","created_at":"2026-05-20T02:05:59.716514+00:00"},{"alias_kind":"pith_short_8","alias_value":"WQPC24LQ","created_at":"2026-05-20T02:05:59.716514+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/WQPC24LQMSJ3INWP4U6IC4ADHO","json":"https://pith.science/pith/WQPC24LQMSJ3INWP4U6IC4ADHO.json","graph_json":"https://pith.science/api/pith-number/WQPC24LQMSJ3INWP4U6IC4ADHO/graph.json","events_json":"https://pith.science/api/pith-number/WQPC24LQMSJ3INWP4U6IC4ADHO/events.json","paper":"https://pith.science/paper/WQPC24LQ"},"agent_actions":{"view_html":"https://pith.science/pith/WQPC24LQMSJ3INWP4U6IC4ADHO","download_json":"https://pith.science/pith/WQPC24LQMSJ3INWP4U6IC4ADHO.json","view_paper":"https://pith.science/paper/WQPC24LQ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.20035&json=true","fetch_graph":"https://pith.science/api/pith-number/WQPC24LQMSJ3INWP4U6IC4ADHO/graph.json","fetch_events":"https://pith.science/api/pith-number/WQPC24LQMSJ3INWP4U6IC4ADHO/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/WQPC24LQMSJ3INWP4U6IC4ADHO/action/timestamp_anchor","attest_storage":"https://pith.science/pith/WQPC24LQMSJ3INWP4U6IC4ADHO/action/storage_attestation","attest_author":"https://pith.science/pith/WQPC24LQMSJ3INWP4U6IC4ADHO/action/author_attestation","sign_citation":"https://pith.science/pith/WQPC24LQMSJ3INWP4U6IC4ADHO/action/citation_signature","submit_replication":"https://pith.science/pith/WQPC24LQMSJ3INWP4U6IC4ADHO/action/replication_record"}},"created_at":"2026-05-20T02:05:59.716514+00:00","updated_at":"2026-05-20T02:05:59.716514+00:00"}