{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:F5SAP2TVLX4NM6K2KGMP5NZQA6","short_pith_number":"pith:F5SAP2TV","canonical_record":{"source":{"id":"2605.14458","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-05-14T06:54:37Z","cross_cats_sorted":[],"title_canon_sha256":"2f0d6d89da70f9c913e7892280f53c6cf2693e18e4a8da795994f65cbc2d4b85","abstract_canon_sha256":"a2aec49115402a41aa1f34d23c264ee1108ba0bcde0b78447ce43214e684e684"},"schema_version":"1.0"},"canonical_sha256":"2f6407ea755df8d6795a5198feb730078f9458cac5c43164d03ff50bdcdbf789","source":{"kind":"arxiv","id":"2605.14458","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.14458","created_at":"2026-05-17T23:39:06Z"},{"alias_kind":"arxiv_version","alias_value":"2605.14458v1","created_at":"2026-05-17T23:39:06Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.14458","created_at":"2026-05-17T23:39:06Z"},{"alias_kind":"pith_short_12","alias_value":"F5SAP2TVLX4N","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"F5SAP2TVLX4NM6K2","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"F5SAP2TV","created_at":"2026-05-18T12:33:37Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:F5SAP2TVLX4NM6K2KGMP5NZQA6","target":"record","payload":{"canonical_record":{"source":{"id":"2605.14458","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-05-14T06:54:37Z","cross_cats_sorted":[],"title_canon_sha256":"2f0d6d89da70f9c913e7892280f53c6cf2693e18e4a8da795994f65cbc2d4b85","abstract_canon_sha256":"a2aec49115402a41aa1f34d23c264ee1108ba0bcde0b78447ce43214e684e684"},"schema_version":"1.0"},"canonical_sha256":"2f6407ea755df8d6795a5198feb730078f9458cac5c43164d03ff50bdcdbf789","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:39:06.814104Z","signature_b64":"yKOZ1M9DmjEkAqVApk9kWYk26IYzdm9cxACpS+tOitTydBOh6fBbGhuJz+e5ddUQKBWOpRNGds77rVCOL362BQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2f6407ea755df8d6795a5198feb730078f9458cac5c43164d03ff50bdcdbf789","last_reissued_at":"2026-05-17T23:39:06.813478Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:39:06.813478Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2605.14458","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-17T23:39:06Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"mFBq5ngh3n8JuJJJ0tZ+ZnFSchk/S3kC6FLUoEMWVSX9n/dhWv4mK8GAH3yL8RdTN5BhNbj74Co7hZsbvqidAQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-27T18:07:49.409140Z"},"content_sha256":"c874ecd400549fb0cd3a91d354755bcfcf9a6251836eb3d1a9eaf1841f921c2e","schema_version":"1.0","event_id":"sha256:c874ecd400549fb0cd3a91d354755bcfcf9a6251836eb3d1a9eaf1841f921c2e"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:F5SAP2TVLX4NM6K2KGMP5NZQA6","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"OmniDrop: Layer-wise Token Pruning for Omni-modal LLMs via Query-Guidance","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Layer-wise token pruning inside the LLM decoder, guided by text queries, allows omni-modal models to process audiovisual inputs faster while maintaining or improving accuracy.","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Hyemi Jang, Jongsun Lee, Jooyoung Choi, Minseo Choi, Yeo Jeong Park, Yongkweon Jeon","submitted_at":"2026-05-14T06:54:37Z","abstract_excerpt":"Omni-modal large language models have demonstrated remarkable potential in holistic multimodal understanding; however, the token explosion caused by high-resolution audio and video inputs remains a critical bottleneck for real-time applications and long-form reasoning. Existing omni-modal token compression methods typically prune tokens at the input embedding level, relying on audio-video similarity or temporal co-occurrence as proxies for semantic relevance. In practice, such assumptions are often unreliable. To address this limitation, we propose OmniDrop, a training-free, layer-wise token p"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Experimental results across various audiovisual benchmarks demonstrate that OmniDrop outperforms all baselines by up to 3.58 points while reducing prefill latency by up to 40% and memory usage by up to 14.7%.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That performing initial fusion in early layers followed by aggressive pruning in deeper layers, guided by text queries, reliably preserves semantic information without task-specific degradation.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"OmniDrop is a training-free layer-wise token pruning framework for omni-modal LLMs that uses query guidance and temporal diversity to reduce prefill latency by up to 40% and memory by 14.7% while improving benchmark scores by up to 3.58 points.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Layer-wise token pruning inside the LLM decoder, guided by text queries, allows omni-modal models to process audiovisual inputs faster while maintaining or improving accuracy.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"8068274d93d673443d4ac5aa68336733ebd366d5454111653e01eb7898750233"},"source":{"id":"2605.14458","kind":"arxiv","version":1},"verdict":{"id":"f678d1cc-701a-46a3-875e-9bf574146f08","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T01:32:16.886455Z","strongest_claim":"Experimental results across various audiovisual benchmarks demonstrate that OmniDrop outperforms all baselines by up to 3.58 points while reducing prefill latency by up to 40% and memory usage by up to 14.7%.","one_line_summary":"OmniDrop is a training-free layer-wise token pruning framework for omni-modal LLMs that uses query guidance and temporal diversity to reduce prefill latency by up to 40% and memory by 14.7% while improving benchmark scores by up to 3.58 points.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That performing initial fusion in early layers followed by aggressive pruning in deeper layers, guided by text queries, reliably preserves semantic information without task-specific degradation.","pith_extraction_headline":"Layer-wise token pruning inside the LLM decoder, guided by text queries, allows omni-modal models to process audiovisual inputs faster while maintaining or improving accuracy."},"references":{"count":33,"sample":[{"doi":"","year":2025,"title":"Divprune: Diversity-based visual token pruning for large multimodal models, 2025","work_id":"7d34fc13-2b41-48af-93a2-a4d706d22a94","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"Token merging: Your vit but faster, 2023","work_id":"af1ae900-f884-4cb0-bebd-85d3b20d6230","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"An image is worth 1/2 tokens after layer 2: Plug-and-play inference acceleration for large vision-language models","work_id":"2c834314-7394-40d4-9cda-458a1172cdf8","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities","work_id":"008df105-2fdd-45d8-857a-8e35868aecb6","ref_index":4,"cited_arxiv_id":"2507.06261","is_internal_anchor":true},{"doi":"","year":2024,"title":"FlashAttention-2: Faster attention with better parallelism and work partitioning","work_id":"3c743fc7-8c9b-4a72-ad8a-26c5a7636140","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":33,"snapshot_sha256":"ac784e2b6b81fe29eccc8ca31a71cb2f97fb0e17b7aad919b3f8ed230aa9fc0a","internal_anchors":5},"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"},"verdict_id":"f678d1cc-701a-46a3-875e-9bf574146f08"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-17T23:39:06Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"DcJ8/waWdchcurUQP6wlaIAkjWtr3eZExP79ObRtAvkFM2Gc8AFlWOy6lxRbvjVz7tCMCiBBfLSPxhwQrrDHBQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-27T18:07:49.409996Z"},"content_sha256":"addb8b432d795d32968e103d93f32f233c393654c4b00442f47d6e32082f2b53","schema_version":"1.0","event_id":"sha256:addb8b432d795d32968e103d93f32f233c393654c4b00442f47d6e32082f2b53"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/F5SAP2TVLX4NM6K2KGMP5NZQA6/bundle.json","state_url":"https://pith.science/pith/F5SAP2TVLX4NM6K2KGMP5NZQA6/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/F5SAP2TVLX4NM6K2KGMP5NZQA6/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-05-27T18:07:49Z","links":{"resolver":"https://pith.science/pith/F5SAP2TVLX4NM6K2KGMP5NZQA6","bundle":"https://pith.science/pith/F5SAP2TVLX4NM6K2KGMP5NZQA6/bundle.json","state":"https://pith.science/pith/F5SAP2TVLX4NM6K2KGMP5NZQA6/state.json","well_known_bundle":"https://pith.science/.well-known/pith/F5SAP2TVLX4NM6K2KGMP5NZQA6/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:F5SAP2TVLX4NM6K2KGMP5NZQA6","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"a2aec49115402a41aa1f34d23c264ee1108ba0bcde0b78447ce43214e684e684","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-05-14T06:54:37Z","title_canon_sha256":"2f0d6d89da70f9c913e7892280f53c6cf2693e18e4a8da795994f65cbc2d4b85"},"schema_version":"1.0","source":{"id":"2605.14458","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.14458","created_at":"2026-05-17T23:39:06Z"},{"alias_kind":"arxiv_version","alias_value":"2605.14458v1","created_at":"2026-05-17T23:39:06Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.14458","created_at":"2026-05-17T23:39:06Z"},{"alias_kind":"pith_short_12","alias_value":"F5SAP2TVLX4N","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"F5SAP2TVLX4NM6K2","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"F5SAP2TV","created_at":"2026-05-18T12:33:37Z"}],"graph_snapshots":[{"event_id":"sha256:addb8b432d795d32968e103d93f32f233c393654c4b00442f47d6e32082f2b53","target":"graph","created_at":"2026-05-17T23:39:06Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"Experimental results across various audiovisual benchmarks demonstrate that OmniDrop outperforms all baselines by up to 3.58 points while reducing prefill latency by up to 40% and memory usage by up to 14.7%."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That performing initial fusion in early layers followed by aggressive pruning in deeper layers, guided by text queries, reliably preserves semantic information without task-specific degradation."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"OmniDrop is a training-free layer-wise token pruning framework for omni-modal LLMs that uses query guidance and temporal diversity to reduce prefill latency by up to 40% and memory by 14.7% while improving benchmark scores by up to 3.58 points."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Layer-wise token pruning inside the LLM decoder, guided by text queries, allows omni-modal models to process audiovisual inputs faster while maintaining or improving accuracy."}],"snapshot_sha256":"8068274d93d673443d4ac5aa68336733ebd366d5454111653e01eb7898750233"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"paper":{"abstract_excerpt":"Omni-modal large language models have demonstrated remarkable potential in holistic multimodal understanding; however, the token explosion caused by high-resolution audio and video inputs remains a critical bottleneck for real-time applications and long-form reasoning. Existing omni-modal token compression methods typically prune tokens at the input embedding level, relying on audio-video similarity or temporal co-occurrence as proxies for semantic relevance. In practice, such assumptions are often unreliable. To address this limitation, we propose OmniDrop, a training-free, layer-wise token p","authors_text":"Hyemi Jang, Jongsun Lee, Jooyoung Choi, Minseo Choi, Yeo Jeong Park, Yongkweon Jeon","cross_cats":[],"headline":"Layer-wise token pruning inside the LLM decoder, guided by text queries, allows omni-modal models to process audiovisual inputs faster while maintaining or improving accuracy.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-05-14T06:54:37Z","title":"OmniDrop: Layer-wise Token Pruning for Omni-modal LLMs via Query-Guidance"},"references":{"count":33,"internal_anchors":5,"resolved_work":33,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"Divprune: Diversity-based visual token pruning for large multimodal models, 2025","work_id":"7d34fc13-2b41-48af-93a2-a4d706d22a94","year":2025},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":2,"title":"Token merging: Your vit but faster, 2023","work_id":"af1ae900-f884-4cb0-bebd-85d3b20d6230","year":2023},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":3,"title":"An image is worth 1/2 tokens after layer 2: Plug-and-play inference acceleration for large vision-language models","work_id":"2c834314-7394-40d4-9cda-458a1172cdf8","year":2024},{"cited_arxiv_id":"2507.06261","doi":"","is_internal_anchor":true,"ref_index":4,"title":"Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities","work_id":"008df105-2fdd-45d8-857a-8e35868aecb6","year":2025},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"FlashAttention-2: Faster attention with better parallelism and work partitioning","work_id":"3c743fc7-8c9b-4a72-ad8a-26c5a7636140","year":2024}],"snapshot_sha256":"ac784e2b6b81fe29eccc8ca31a71cb2f97fb0e17b7aad919b3f8ed230aa9fc0a"},"source":{"id":"2605.14458","kind":"arxiv","version":1},"verdict":{"created_at":"2026-05-15T01:32:16.886455Z","id":"f678d1cc-701a-46a3-875e-9bf574146f08","model_set":{"reader":"grok-4.3"},"one_line_summary":"OmniDrop is a training-free layer-wise token pruning framework for omni-modal LLMs that uses query guidance and temporal diversity to reduce prefill latency by up to 40% and memory by 14.7% while improving benchmark scores by up to 3.58 points.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Layer-wise token pruning inside the LLM decoder, guided by text queries, allows omni-modal models to process audiovisual inputs faster while maintaining or improving accuracy.","strongest_claim":"Experimental results across various audiovisual benchmarks demonstrate that OmniDrop outperforms all baselines by up to 3.58 points while reducing prefill latency by up to 40% and memory usage by up to 14.7%.","weakest_assumption":"That performing initial fusion in early layers followed by aggressive pruning in deeper layers, guided by text queries, reliably preserves semantic information without task-specific degradation."}},"verdict_id":"f678d1cc-701a-46a3-875e-9bf574146f08"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:c874ecd400549fb0cd3a91d354755bcfcf9a6251836eb3d1a9eaf1841f921c2e","target":"record","created_at":"2026-05-17T23:39:06Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"a2aec49115402a41aa1f34d23c264ee1108ba0bcde0b78447ce43214e684e684","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-05-14T06:54:37Z","title_canon_sha256":"2f0d6d89da70f9c913e7892280f53c6cf2693e18e4a8da795994f65cbc2d4b85"},"schema_version":"1.0","source":{"id":"2605.14458","kind":"arxiv","version":1}},"canonical_sha256":"2f6407ea755df8d6795a5198feb730078f9458cac5c43164d03ff50bdcdbf789","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"2f6407ea755df8d6795a5198feb730078f9458cac5c43164d03ff50bdcdbf789","first_computed_at":"2026-05-17T23:39:06.813478Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-17T23:39:06.813478Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"yKOZ1M9DmjEkAqVApk9kWYk26IYzdm9cxACpS+tOitTydBOh6fBbGhuJz+e5ddUQKBWOpRNGds77rVCOL362BQ==","signature_status":"signed_v1","signed_at":"2026-05-17T23:39:06.814104Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.14458","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:c874ecd400549fb0cd3a91d354755bcfcf9a6251836eb3d1a9eaf1841f921c2e","sha256:addb8b432d795d32968e103d93f32f233c393654c4b00442f47d6e32082f2b53"],"state_sha256":"4ba094bec91ba712f5846d00a80832f003b74fe136c432a278a3adbfeb3dd00e"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"3YZQMtSmlyZKIJM+no1NeXsNWDMzq55Rd4A/sw3j8TB9Wt91Lkix0vJigxYKZGGBwggf2uxQgn+kNoUhLBp/Cw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-05-27T18:07:49.413425Z","bundle_sha256":"7965064961cd38b044e7689af4ef49b439a49bc5d3a83f39dbe4ed1e74d6074a"}}