{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:N2BCLDNRQUPXC3XBBARF4FM5AY","short_pith_number":"pith:N2BCLDNR","schema_version":"1.0","canonical_sha256":"6e82258db1851f716ee108225e159d0618944b6ea0d6710791de18d8d1a4266c","source":{"kind":"arxiv","id":"2608.08794","version":1},"attestation_state":"computed","paper":{"title":"Deferred Audio Pruning with Local Audio-Visual Dynamics for Omni-LLMs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.MM","cs.SD"],"primary_cat":"cs.AI","authors_text":"Hongyeob Kim, Kyeongyoon Lee, Sungeun Hong, Youngeun Kim","submitted_at":"2026-08-09T16:18:41Z","abstract_excerpt":"Omni-modal LLMs jointly process audio, video, and text, but long multimodal sequences incur substantial prefill and KV-cache costs. Existing omni-modal compression methods primarily focus on pre-LLM token reduction, leaving modality-specific compression across the LLM boundary underexplored. We propose A-PACK, a two-stage framework that defers audio pruning until query-conditioned multimodal interactions emerge. Our analysis shows that audio exhibits higher task-relevant information density and representational diversity per token than video. We further find that local audio-visual dynamics pr"},"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":"2608.08794","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2026-08-09T16:18:41Z","cross_cats_sorted":["cs.MM","cs.SD"],"title_canon_sha256":"7ba7e89c071f7f2972a135e61b7431255532c3c02c63c70099f53aac39b6cf27","abstract_canon_sha256":"b88b01957a2f9ef62321f5627423d4800220f3e2bca3ae4d432e463dd6f8d906"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-11T01:25:21.568034Z","signature_b64":"UGE45tch2O1RSfL9fWtWTLQSyPehu2Q21yTm53/DO99SKsJnExNjV8Zl7pE5zlTcO6uO/KYGfSBNI4PrXhZaDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6e82258db1851f716ee108225e159d0618944b6ea0d6710791de18d8d1a4266c","last_reissued_at":"2026-08-11T01:25:21.565388Z","signature_status":"signed_v1","first_computed_at":"2026-08-11T01:25:21.565388Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Deferred Audio Pruning with Local Audio-Visual Dynamics for Omni-LLMs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.MM","cs.SD"],"primary_cat":"cs.AI","authors_text":"Hongyeob Kim, Kyeongyoon Lee, Sungeun Hong, Youngeun Kim","submitted_at":"2026-08-09T16:18:41Z","abstract_excerpt":"Omni-modal LLMs jointly process audio, video, and text, but long multimodal sequences incur substantial prefill and KV-cache costs. Existing omni-modal compression methods primarily focus on pre-LLM token reduction, leaving modality-specific compression across the LLM boundary underexplored. We propose A-PACK, a two-stage framework that defers audio pruning until query-conditioned multimodal interactions emerge. Our analysis shows that audio exhibits higher task-relevant information density and representational diversity per token than video. We further find that local audio-visual dynamics pr"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2608.08794","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/2608.08794/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":"2608.08794","created_at":"2026-08-11T01:25:21.566278+00:00"},{"alias_kind":"arxiv_version","alias_value":"2608.08794v1","created_at":"2026-08-11T01:25:21.566278+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2608.08794","created_at":"2026-08-11T01:25:21.566278+00:00"},{"alias_kind":"pith_short_12","alias_value":"N2BCLDNRQUPX","created_at":"2026-08-11T01:25:21.566278+00:00"},{"alias_kind":"pith_short_16","alias_value":"N2BCLDNRQUPXC3XB","created_at":"2026-08-11T01:25:21.566278+00:00"},{"alias_kind":"pith_short_8","alias_value":"N2BCLDNR","created_at":"2026-08-11T01:25:21.566278+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/N2BCLDNRQUPXC3XBBARF4FM5AY","json":"https://pith.science/pith/N2BCLDNRQUPXC3XBBARF4FM5AY.json","graph_json":"https://pith.science/api/pith-number/N2BCLDNRQUPXC3XBBARF4FM5AY/graph.json","events_json":"https://pith.science/api/pith-number/N2BCLDNRQUPXC3XBBARF4FM5AY/events.json","paper":"https://pith.science/paper/N2BCLDNR"},"agent_actions":{"view_html":"https://pith.science/pith/N2BCLDNRQUPXC3XBBARF4FM5AY","download_json":"https://pith.science/pith/N2BCLDNRQUPXC3XBBARF4FM5AY.json","view_paper":"https://pith.science/paper/N2BCLDNR","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2608.08794&json=true","fetch_graph":"https://pith.science/api/pith-number/N2BCLDNRQUPXC3XBBARF4FM5AY/graph.json","fetch_events":"https://pith.science/api/pith-number/N2BCLDNRQUPXC3XBBARF4FM5AY/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/N2BCLDNRQUPXC3XBBARF4FM5AY/action/timestamp_anchor","attest_storage":"https://pith.science/pith/N2BCLDNRQUPXC3XBBARF4FM5AY/action/storage_attestation","attest_author":"https://pith.science/pith/N2BCLDNRQUPXC3XBBARF4FM5AY/action/author_attestation","sign_citation":"https://pith.science/pith/N2BCLDNRQUPXC3XBBARF4FM5AY/action/citation_signature","submit_replication":"https://pith.science/pith/N2BCLDNRQUPXC3XBBARF4FM5AY/action/replication_record"}},"created_at":"2026-08-11T01:25:21.566278+00:00","updated_at":"2026-08-11T01:25:21.566278+00:00"}