{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:H2OIZBHNFCOWRFZW4Y7BC2YRF5","short_pith_number":"pith:H2OIZBHN","schema_version":"1.0","canonical_sha256":"3e9c8c84ed289d689736e63e116b112f417a2271f30077f58706a07b52d99ffc","source":{"kind":"arxiv","id":"2508.01548","version":1},"attestation_state":"computed","paper":{"title":"A Glimpse to Compress: Dynamic Visual Token Pruning for Large Vision-Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Ming-Ming Cheng, Peng-Tao Jiang, Qibin Hou, Qilong Wang, Quan-Sheng Zeng, Yunheng Li, Zuxuan Wu","submitted_at":"2025-08-03T02:15:43Z","abstract_excerpt":"Visual token compression is critical for Large Vision-Language Models (LVLMs) to efficiently process high-resolution inputs. Existing methods that typically adopt fixed compression ratios cannot adapt to scenes of varying complexity, often causing imprecise pruning that discards informative visual tokens and results in degraded model performance. To address this issue, we introduce a dynamic pruning framework, GlimpsePrune, inspired by human cognition. It takes a data-driven ''glimpse'' and prunes irrelevant visual tokens in a single forward pass before answer generation. This approach prunes "},"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":"2508.01548","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2025-08-03T02:15:43Z","cross_cats_sorted":[],"title_canon_sha256":"52df525c70f002cf6fea61834d184b321d5c75d2811072973cb9fe3ab5253f42","abstract_canon_sha256":"1c8d14b66f7f5f7b1c758ac386b2fe67873b397c1112a08c2b7a5abfe4b6554a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:47:40.947703Z","signature_b64":"/r9EYUVsrN+o72djv+oDUQ8eJphW7LW90mc3eDJdo7OmxMgpxrHXeilcdGInJJsURjX/dh9xKY3rLqyypsFoDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3e9c8c84ed289d689736e63e116b112f417a2271f30077f58706a07b52d99ffc","last_reissued_at":"2026-07-05T11:47:40.947242Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:47:40.947242Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"A Glimpse to Compress: Dynamic Visual Token Pruning for Large Vision-Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Ming-Ming Cheng, Peng-Tao Jiang, Qibin Hou, Qilong Wang, Quan-Sheng Zeng, Yunheng Li, Zuxuan Wu","submitted_at":"2025-08-03T02:15:43Z","abstract_excerpt":"Visual token compression is critical for Large Vision-Language Models (LVLMs) to efficiently process high-resolution inputs. Existing methods that typically adopt fixed compression ratios cannot adapt to scenes of varying complexity, often causing imprecise pruning that discards informative visual tokens and results in degraded model performance. To address this issue, we introduce a dynamic pruning framework, GlimpsePrune, inspired by human cognition. It takes a data-driven ''glimpse'' and prunes irrelevant visual tokens in a single forward pass before answer generation. This approach prunes "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2508.01548","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/2508.01548/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":"2508.01548","created_at":"2026-07-05T11:47:40.947299+00:00"},{"alias_kind":"arxiv_version","alias_value":"2508.01548v1","created_at":"2026-07-05T11:47:40.947299+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2508.01548","created_at":"2026-07-05T11:47:40.947299+00:00"},{"alias_kind":"pith_short_12","alias_value":"H2OIZBHNFCOW","created_at":"2026-07-05T11:47:40.947299+00:00"},{"alias_kind":"pith_short_16","alias_value":"H2OIZBHNFCOWRFZW","created_at":"2026-07-05T11:47:40.947299+00:00"},{"alias_kind":"pith_short_8","alias_value":"H2OIZBHN","created_at":"2026-07-05T11:47:40.947299+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.24165","citing_title":"Spectral Evolution-Guided Token Pruning in Multimodal Large Language Models","ref_index":85,"is_internal_anchor":false},{"citing_arxiv_id":"2606.24156","citing_title":"Accelerating Multimodal Large Language Models with Prior-Corrected Token Reduction","ref_index":86,"is_internal_anchor":false},{"citing_arxiv_id":"2605.28051","citing_title":"Beyond Surrogate Gradients: Fully Differentiable Token Pruning for Vision-Language Models","ref_index":53,"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":93,"is_internal_anchor":false},{"citing_arxiv_id":"2604.11122","citing_title":"Semantic-Geometric Dual Compression: Training-Free Visual Token Reduction for Ultra-High-Resolution Remote Sensing Understanding","ref_index":53,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/H2OIZBHNFCOWRFZW4Y7BC2YRF5","json":"https://pith.science/pith/H2OIZBHNFCOWRFZW4Y7BC2YRF5.json","graph_json":"https://pith.science/api/pith-number/H2OIZBHNFCOWRFZW4Y7BC2YRF5/graph.json","events_json":"https://pith.science/api/pith-number/H2OIZBHNFCOWRFZW4Y7BC2YRF5/events.json","paper":"https://pith.science/paper/H2OIZBHN"},"agent_actions":{"view_html":"https://pith.science/pith/H2OIZBHNFCOWRFZW4Y7BC2YRF5","download_json":"https://pith.science/pith/H2OIZBHNFCOWRFZW4Y7BC2YRF5.json","view_paper":"https://pith.science/paper/H2OIZBHN","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2508.01548&json=true","fetch_graph":"https://pith.science/api/pith-number/H2OIZBHNFCOWRFZW4Y7BC2YRF5/graph.json","fetch_events":"https://pith.science/api/pith-number/H2OIZBHNFCOWRFZW4Y7BC2YRF5/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/H2OIZBHNFCOWRFZW4Y7BC2YRF5/action/timestamp_anchor","attest_storage":"https://pith.science/pith/H2OIZBHNFCOWRFZW4Y7BC2YRF5/action/storage_attestation","attest_author":"https://pith.science/pith/H2OIZBHNFCOWRFZW4Y7BC2YRF5/action/author_attestation","sign_citation":"https://pith.science/pith/H2OIZBHNFCOWRFZW4Y7BC2YRF5/action/citation_signature","submit_replication":"https://pith.science/pith/H2OIZBHNFCOWRFZW4Y7BC2YRF5/action/replication_record"}},"created_at":"2026-07-05T11:47:40.947299+00:00","updated_at":"2026-07-05T11:47:40.947299+00:00"}