{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:LWZ6PLF3WNKIGOBDQ4BCWAAS5N","short_pith_number":"pith:LWZ6PLF3","schema_version":"1.0","canonical_sha256":"5db3e7acbbb35483382387022b0012eb5be66807d8a47934c2641c17e0b7b0de","source":{"kind":"arxiv","id":"2608.04496","version":1},"attestation_state":"computed","paper":{"title":"DIVE: Dynamic Iterative Visual Evidence Construction for Efficient Vision-Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CV","authors_text":"Chen Zhong, Guangyi Yang, Jiepan Li, Wei He, Xiao An, Zijie Wang","submitted_at":"2026-08-05T06:32:25Z","abstract_excerpt":"Visual inputs in vision-language models (VLMs) are often encoded into substantially longer token sequences than text, making visual tokens a major bottleneck for efficient inference. Abundant recent methods address this bottleneck by scoring token importance and pruning low-scoring tokens in a single pass. However, one-shot scoring is insufficient because a token's prompt-relevant usefulness depends on the evidence already retained. Motivated by this insight, we introduce DIVE (Dynamic Iterative Visual Evidence Construction), a training-free framework that recasts visual-token pruning as dynam"},"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":"2608.04496","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-08-05T06:32:25Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"abb1979733dbe48e6c9322cc58ac4bc44c2c31973f6f553ede9613e4635e6803","abstract_canon_sha256":"88ed1e84f3bcb58687cf04f4079fe1c179668539b4cb1409199970239e037529"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-06T01:35:42.858835Z","signature_b64":"uEIHHnUUKnA/ANj4RsXNFxWXPywLED1soBoo5pq8IlHFFn3awBblbRdcBMsVhYEdGK33yEJr96qcXhhhfvhXDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5db3e7acbbb35483382387022b0012eb5be66807d8a47934c2641c17e0b7b0de","last_reissued_at":"2026-08-06T01:35:42.857456Z","signature_status":"signed_v1","first_computed_at":"2026-08-06T01:35:42.857456Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"DIVE: Dynamic Iterative Visual Evidence Construction for Efficient Vision-Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CV","authors_text":"Chen Zhong, Guangyi Yang, Jiepan Li, Wei He, Xiao An, Zijie Wang","submitted_at":"2026-08-05T06:32:25Z","abstract_excerpt":"Visual inputs in vision-language models (VLMs) are often encoded into substantially longer token sequences than text, making visual tokens a major bottleneck for efficient inference. Abundant recent methods address this bottleneck by scoring token importance and pruning low-scoring tokens in a single pass. However, one-shot scoring is insufficient because a token's prompt-relevant usefulness depends on the evidence already retained. Motivated by this insight, we introduce DIVE (Dynamic Iterative Visual Evidence Construction), a training-free framework that recasts visual-token pruning as dynam"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2608.04496","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/2608.04496/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":"2608.04496","created_at":"2026-08-06T01:35:42.859231+00:00"},{"alias_kind":"arxiv_version","alias_value":"2608.04496v1","created_at":"2026-08-06T01:35:42.859231+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2608.04496","created_at":"2026-08-06T01:35:42.859231+00:00"},{"alias_kind":"pith_short_12","alias_value":"LWZ6PLF3WNKI","created_at":"2026-08-06T01:35:42.859231+00:00"},{"alias_kind":"pith_short_16","alias_value":"LWZ6PLF3WNKIGOBD","created_at":"2026-08-06T01:35:42.859231+00:00"},{"alias_kind":"pith_short_8","alias_value":"LWZ6PLF3","created_at":"2026-08-06T01:35:42.859231+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/LWZ6PLF3WNKIGOBDQ4BCWAAS5N","json":"https://pith.science/pith/LWZ6PLF3WNKIGOBDQ4BCWAAS5N.json","graph_json":"https://pith.science/api/pith-number/LWZ6PLF3WNKIGOBDQ4BCWAAS5N/graph.json","events_json":"https://pith.science/api/pith-number/LWZ6PLF3WNKIGOBDQ4BCWAAS5N/events.json","paper":"https://pith.science/paper/LWZ6PLF3"},"agent_actions":{"view_html":"https://pith.science/pith/LWZ6PLF3WNKIGOBDQ4BCWAAS5N","download_json":"https://pith.science/pith/LWZ6PLF3WNKIGOBDQ4BCWAAS5N.json","view_paper":"https://pith.science/paper/LWZ6PLF3","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2608.04496&json=true","fetch_graph":"https://pith.science/api/pith-number/LWZ6PLF3WNKIGOBDQ4BCWAAS5N/graph.json","fetch_events":"https://pith.science/api/pith-number/LWZ6PLF3WNKIGOBDQ4BCWAAS5N/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/LWZ6PLF3WNKIGOBDQ4BCWAAS5N/action/timestamp_anchor","attest_storage":"https://pith.science/pith/LWZ6PLF3WNKIGOBDQ4BCWAAS5N/action/storage_attestation","attest_author":"https://pith.science/pith/LWZ6PLF3WNKIGOBDQ4BCWAAS5N/action/author_attestation","sign_citation":"https://pith.science/pith/LWZ6PLF3WNKIGOBDQ4BCWAAS5N/action/citation_signature","submit_replication":"https://pith.science/pith/LWZ6PLF3WNKIGOBDQ4BCWAAS5N/action/replication_record"}},"created_at":"2026-08-06T01:35:42.859231+00:00","updated_at":"2026-08-06T01:35:42.859231+00:00"}