{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2024:MHQFABVDPDJ25UCI6FPPO2QXTD","short_pith_number":"pith:MHQFABVD","canonical_record":{"source":{"id":"2410.04417","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CV","submitted_at":"2024-10-06T09:18:04Z","cross_cats_sorted":[],"title_canon_sha256":"5a5b188732cf551e1c8e59a062df3da8b29f8ba0ba1d9a822d2800cd56f8afd6","abstract_canon_sha256":"7f4d6f8fdd0d8c4f7bdd63f53aeab9330445dea86771847def61af753e0e9484"},"schema_version":"1.0"},"canonical_sha256":"61e05006a378d3aed048f15ef76a1798fae4e7f6a4698c1b62fcea3962ec2680","source":{"kind":"arxiv","id":"2410.04417","version":4},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2410.04417","created_at":"2026-05-17T23:38:52Z"},{"alias_kind":"arxiv_version","alias_value":"2410.04417v4","created_at":"2026-05-17T23:38:52Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.04417","created_at":"2026-05-17T23:38:52Z"},{"alias_kind":"pith_short_12","alias_value":"MHQFABVDPDJ2","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"MHQFABVDPDJ25UCI","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"MHQFABVD","created_at":"2026-05-18T12:33:37Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2024:MHQFABVDPDJ25UCI6FPPO2QXTD","target":"record","payload":{"canonical_record":{"source":{"id":"2410.04417","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CV","submitted_at":"2024-10-06T09:18:04Z","cross_cats_sorted":[],"title_canon_sha256":"5a5b188732cf551e1c8e59a062df3da8b29f8ba0ba1d9a822d2800cd56f8afd6","abstract_canon_sha256":"7f4d6f8fdd0d8c4f7bdd63f53aeab9330445dea86771847def61af753e0e9484"},"schema_version":"1.0"},"canonical_sha256":"61e05006a378d3aed048f15ef76a1798fae4e7f6a4698c1b62fcea3962ec2680","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:38:52.301150Z","signature_b64":"BJKAVmjjmJXhZh+KVg5lC4AFZJ1JX19hVGUcVg3BuWtXjD7P6DwPGIATXGKVlvCAlk+y9zQqxUMXXU8ZVu1+Cw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"61e05006a378d3aed048f15ef76a1798fae4e7f6a4698c1b62fcea3962ec2680","last_reissued_at":"2026-05-17T23:38:52.300692Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:38:52.300692Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2410.04417","source_version":4,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-17T23:38:52Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"n7rxb2Xa7yGszoyeuXtHQGLwQ/W1wGFsgSiDo3mqD1b89xLCVmHqxnm5TCf2QN7OD37eKlHXa6N+edodu5joBw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-25T19:33:03.001212Z"},"content_sha256":"3eb97bcb1a293f9d9ed46f2d3e585e4c1faa487783a08e7b88a5e857e2d6b35a","schema_version":"1.0","event_id":"sha256:3eb97bcb1a293f9d9ed46f2d3e585e4c1faa487783a08e7b88a5e857e2d6b35a"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2024:MHQFABVDPDJ25UCI6FPPO2QXTD","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"SparseVLM: Visual Token Sparsification for Efficient Vision-Language Model Inference","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"SparseVLM prunes visual tokens in VLMs using text attention scores without any training or added parameters.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Chun-Kai Fan, Denis Gudovskiy, Junpeng Ma, Kuan Cheng, Kurt Keutzer, Shanghang Zhang, Tao Huang, Tomoyuki Okuno, Wenzhao Zheng, Yohei Nakata, Yuan Zhang","submitted_at":"2024-10-06T09:18:04Z","abstract_excerpt":"In vision-language models (VLMs), visual tokens usually bear a significant amount of computational overhead despite sparsity of information in them when compared to text tokens. To address this, most existing methods learn a network to prune redundant visual tokens using certain training data. Differently, we propose a text-guided training-free token optimization mechanism dubbed SparseVLM that eliminates the need of extra parameters or fine-tuning costs. Given that visual tokens complement text tokens in VLM's linguistic reasoning, we select relevant text tokens to rate the significance of vi"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"SparseVLM increases the efficiency of various VLMs in a number of image and video understanding tasks. For example, LLaVA when equipped with SparseVLM achieves 54% reduction in FLOPs, 37% decrease in CUDA latency while maintaining 97% of its original accuracy.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That self-attention scores between selected text tokens and visual tokens reliably identify which visual tokens can be pruned or recycled without losing task-critical information.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"SparseVLM uses text-guided attention to prune and recycle visual tokens in VLMs, delivering 54% FLOPs reduction and 37% lower latency with 97% accuracy retention on LLaVA.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"SparseVLM prunes visual tokens in VLMs using text attention scores without any training or added parameters.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"f5f7695a3f895ec4f477af7edd6fda775677f5cf825fee44489a8d8b021ab3d8"},"source":{"id":"2410.04417","kind":"arxiv","version":4},"verdict":{"id":"6bba3ee3-d597-404a-8061-2f18b500a1d1","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T14:51:52.984961Z","strongest_claim":"SparseVLM increases the efficiency of various VLMs in a number of image and video understanding tasks. For example, LLaVA when equipped with SparseVLM achieves 54% reduction in FLOPs, 37% decrease in CUDA latency while maintaining 97% of its original accuracy.","one_line_summary":"SparseVLM uses text-guided attention to prune and recycle visual tokens in VLMs, delivering 54% FLOPs reduction and 37% lower latency with 97% accuracy retention on LLaVA.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That self-attention scores between selected text tokens and visual tokens reliably identify which visual tokens can be pruned or recycled without losing task-critical information.","pith_extraction_headline":"SparseVLM prunes visual tokens in VLMs using text attention scores without any training or added parameters."},"references":{"count":113,"sample":[{"doi":"","year":2022,"title":"Flamingo: a visual language model for few-shot learning","work_id":"fa381b5e-5991-4bbf-a49a-2e90138d814f","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"Qwen-VL: A Versatile Vision-Language Model for Understanding, Localization, Text Reading, and Beyond","work_id":"cbc2bb21-b6bb-46c0-80bf-107e195ffe10","ref_index":2,"cited_arxiv_id":"2308.12966","is_internal_anchor":true},{"doi":"","year":2023,"title":"Token merging: Your vit but faster","work_id":"671514fa-294d-462f-b530-bd25a0b38101","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2020,"title":"D., Dhariwal, P., Neelakantan, A., Shyam, P., Sastry, G., Askell, A., et al","work_id":"9c4d1f20-8715-4aaa-9bb1-9420a625a69d","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"Cai, M., Yang, J., Gao, J., and Lee, Y. J. Matryoshka multimodal models. In International Conference on Learning Representations, 2025","work_id":"9dedb931-d4a0-4bbe-8674-ef4b4d38a5ed","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":113,"snapshot_sha256":"e1a38f11346fc40939c1883d42f7127bb0b240b3dbe26bec2e7c50b547b9be06","internal_anchors":12},"formal_canon":{"evidence_count":2,"snapshot_sha256":"3e88488bc8d9ffba410b08778160f2c70a843dff80be28da54870cd7ca6e3d1d"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":"6bba3ee3-d597-404a-8061-2f18b500a1d1"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-17T23:38:52Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"XA6W2W8WHMHqYMYRGBGiy0WPrcEZ5cFYTU7dFBBied6KhpHsCISnHrg7wKMOZ0yklrjZD1CJbLPZUKhMy6DKDw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-25T19:33:03.002249Z"},"content_sha256":"32aa8b743a23fc656f8668d46921c9e1ad7ca75b7bcb6015179f1bdf959e49e9","schema_version":"1.0","event_id":"sha256:32aa8b743a23fc656f8668d46921c9e1ad7ca75b7bcb6015179f1bdf959e49e9"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/MHQFABVDPDJ25UCI6FPPO2QXTD/bundle.json","state_url":"https://pith.science/pith/MHQFABVDPDJ25UCI6FPPO2QXTD/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/MHQFABVDPDJ25UCI6FPPO2QXTD/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":"2026-05-25T19:33:03Z","links":{"resolver":"https://pith.science/pith/MHQFABVDPDJ25UCI6FPPO2QXTD","bundle":"https://pith.science/pith/MHQFABVDPDJ25UCI6FPPO2QXTD/bundle.json","state":"https://pith.science/pith/MHQFABVDPDJ25UCI6FPPO2QXTD/state.json","well_known_bundle":"https://pith.science/.well-known/pith/MHQFABVDPDJ25UCI6FPPO2QXTD/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:MHQFABVDPDJ25UCI6FPPO2QXTD","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":"7f4d6f8fdd0d8c4f7bdd63f53aeab9330445dea86771847def61af753e0e9484","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CV","submitted_at":"2024-10-06T09:18:04Z","title_canon_sha256":"5a5b188732cf551e1c8e59a062df3da8b29f8ba0ba1d9a822d2800cd56f8afd6"},"schema_version":"1.0","source":{"id":"2410.04417","kind":"arxiv","version":4}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2410.04417","created_at":"2026-05-17T23:38:52Z"},{"alias_kind":"arxiv_version","alias_value":"2410.04417v4","created_at":"2026-05-17T23:38:52Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.04417","created_at":"2026-05-17T23:38:52Z"},{"alias_kind":"pith_short_12","alias_value":"MHQFABVDPDJ2","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"MHQFABVDPDJ25UCI","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"MHQFABVD","created_at":"2026-05-18T12:33:37Z"}],"graph_snapshots":[{"event_id":"sha256:32aa8b743a23fc656f8668d46921c9e1ad7ca75b7bcb6015179f1bdf959e49e9","target":"graph","created_at":"2026-05-17T23:38:52Z","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":"SparseVLM increases the efficiency of various VLMs in a number of image and video understanding tasks. For example, LLaVA when equipped with SparseVLM achieves 54% reduction in FLOPs, 37% decrease in CUDA latency while maintaining 97% of its original accuracy."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That self-attention scores between selected text tokens and visual tokens reliably identify which visual tokens can be pruned or recycled without losing task-critical information."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"SparseVLM uses text-guided attention to prune and recycle visual tokens in VLMs, delivering 54% FLOPs reduction and 37% lower latency with 97% accuracy retention on LLaVA."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"SparseVLM prunes visual tokens in VLMs using text attention scores without any training or added parameters."}],"snapshot_sha256":"f5f7695a3f895ec4f477af7edd6fda775677f5cf825fee44489a8d8b021ab3d8"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"3e88488bc8d9ffba410b08778160f2c70a843dff80be28da54870cd7ca6e3d1d"},"paper":{"abstract_excerpt":"In vision-language models (VLMs), visual tokens usually bear a significant amount of computational overhead despite sparsity of information in them when compared to text tokens. To address this, most existing methods learn a network to prune redundant visual tokens using certain training data. Differently, we propose a text-guided training-free token optimization mechanism dubbed SparseVLM that eliminates the need of extra parameters or fine-tuning costs. Given that visual tokens complement text tokens in VLM's linguistic reasoning, we select relevant text tokens to rate the significance of vi","authors_text":"Chun-Kai Fan, Denis Gudovskiy, Junpeng Ma, Kuan Cheng, Kurt Keutzer, Shanghang Zhang, Tao Huang, Tomoyuki Okuno, Wenzhao Zheng, Yohei Nakata, Yuan Zhang","cross_cats":[],"headline":"SparseVLM prunes visual tokens in VLMs using text attention scores without any training or added parameters.","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CV","submitted_at":"2024-10-06T09:18:04Z","title":"SparseVLM: Visual Token Sparsification for Efficient Vision-Language Model Inference"},"references":{"count":113,"internal_anchors":12,"resolved_work":113,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"Flamingo: a visual language model for few-shot learning","work_id":"fa381b5e-5991-4bbf-a49a-2e90138d814f","year":2022},{"cited_arxiv_id":"2308.12966","doi":"","is_internal_anchor":true,"ref_index":2,"title":"Qwen-VL: A Versatile Vision-Language Model for Understanding, Localization, Text Reading, and Beyond","work_id":"cbc2bb21-b6bb-46c0-80bf-107e195ffe10","year":2023},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":3,"title":"Token merging: Your vit but faster","work_id":"671514fa-294d-462f-b530-bd25a0b38101","year":2023},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"D., Dhariwal, P., Neelakantan, A., Shyam, P., Sastry, G., Askell, A., et al","work_id":"9c4d1f20-8715-4aaa-9bb1-9420a625a69d","year":2020},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"Cai, M., Yang, J., Gao, J., and Lee, Y. J. Matryoshka multimodal models. In International Conference on Learning Representations, 2025","work_id":"9dedb931-d4a0-4bbe-8674-ef4b4d38a5ed","year":2025}],"snapshot_sha256":"e1a38f11346fc40939c1883d42f7127bb0b240b3dbe26bec2e7c50b547b9be06"},"source":{"id":"2410.04417","kind":"arxiv","version":4},"verdict":{"created_at":"2026-05-15T14:51:52.984961Z","id":"6bba3ee3-d597-404a-8061-2f18b500a1d1","model_set":{"reader":"grok-4.3"},"one_line_summary":"SparseVLM uses text-guided attention to prune and recycle visual tokens in VLMs, delivering 54% FLOPs reduction and 37% lower latency with 97% accuracy retention on LLaVA.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"SparseVLM prunes visual tokens in VLMs using text attention scores without any training or added parameters.","strongest_claim":"SparseVLM increases the efficiency of various VLMs in a number of image and video understanding tasks. For example, LLaVA when equipped with SparseVLM achieves 54% reduction in FLOPs, 37% decrease in CUDA latency while maintaining 97% of its original accuracy.","weakest_assumption":"That self-attention scores between selected text tokens and visual tokens reliably identify which visual tokens can be pruned or recycled without losing task-critical information."}},"verdict_id":"6bba3ee3-d597-404a-8061-2f18b500a1d1"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:3eb97bcb1a293f9d9ed46f2d3e585e4c1faa487783a08e7b88a5e857e2d6b35a","target":"record","created_at":"2026-05-17T23:38:52Z","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":"7f4d6f8fdd0d8c4f7bdd63f53aeab9330445dea86771847def61af753e0e9484","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CV","submitted_at":"2024-10-06T09:18:04Z","title_canon_sha256":"5a5b188732cf551e1c8e59a062df3da8b29f8ba0ba1d9a822d2800cd56f8afd6"},"schema_version":"1.0","source":{"id":"2410.04417","kind":"arxiv","version":4}},"canonical_sha256":"61e05006a378d3aed048f15ef76a1798fae4e7f6a4698c1b62fcea3962ec2680","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"61e05006a378d3aed048f15ef76a1798fae4e7f6a4698c1b62fcea3962ec2680","first_computed_at":"2026-05-17T23:38:52.300692Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-17T23:38:52.300692Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"BJKAVmjjmJXhZh+KVg5lC4AFZJ1JX19hVGUcVg3BuWtXjD7P6DwPGIATXGKVlvCAlk+y9zQqxUMXXU8ZVu1+Cw==","signature_status":"signed_v1","signed_at":"2026-05-17T23:38:52.301150Z","signed_message":"canonical_sha256_bytes"},"source_id":"2410.04417","source_kind":"arxiv","source_version":4}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:3eb97bcb1a293f9d9ed46f2d3e585e4c1faa487783a08e7b88a5e857e2d6b35a","sha256:32aa8b743a23fc656f8668d46921c9e1ad7ca75b7bcb6015179f1bdf959e49e9"],"state_sha256":"53fbbf6285d6974c9cd36575b4f47641d0dafce94036d54d19a2c0eac9b7cfa6"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"MMUYwVDNS8q+lpQyjOvL52m0Rh7ttr0f6+S4onYvrwT1d6PsqPYBePFw+TdmnKfcMex5gzAhcdVJ5uL16WrNDA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-05-25T19:33:03.006837Z","bundle_sha256":"46b855564046f0bbd06641220bec8980ee01d25894117773e301b918ec85c1df"}}