{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:ZXO4C6FBFJSD3BVNIBU3AMAPWS","short_pith_number":"pith:ZXO4C6FB","schema_version":"1.0","canonical_sha256":"cdddc178a12a643d86ad4069b0300fb4a3252ef1400f9c3816936ca26cf7a1a4","source":{"kind":"arxiv","id":"2506.03643","version":2},"attestation_state":"computed","paper":{"title":"Images are Worth Variable Length of Representations","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Lingjun Mao, Rodolfo Corona, Wenhao Yan, Xin Liang, Zineng Tang","submitted_at":"2025-06-04T07:40:33Z","abstract_excerpt":"Most existing vision encoders map images into a fixed-length sequence of tokens, overlooking the fact that different images contain varying amounts of information. For example, a visually complex image (e.g., a cluttered room) inherently carries more information and thus deserves more tokens than a simple image (e.g., a blank wall). To address this inefficiency, we propose DOVE, a dynamic vision encoder that produces a variable number of visual tokens (i.e., continuous representation vectors) to reconstruct each image. Our results show that DOVE significantly reduces the average number of toke"},"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":"2506.03643","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2025-06-04T07:40:33Z","cross_cats_sorted":[],"title_canon_sha256":"a011b39cc911cbaddc8759ea2feb96ee6d1c7c1af453de8cdaed412bb921c4b3","abstract_canon_sha256":"ee55d9613d6983a58724930b669305f00870f4bb095004c0ee4bd55b327af306"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:16:36.177923Z","signature_b64":"ZTzlpgI4EcgsWpnTrfUVEdQ193k3Nm9nPGWKjP7ly/B6/BaAGtIUXf/dPxq3geixuq8FrwsfvnwGh5DcBRJCBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"cdddc178a12a643d86ad4069b0300fb4a3252ef1400f9c3816936ca26cf7a1a4","last_reissued_at":"2026-07-05T11:16:36.177390Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:16:36.177390Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Images are Worth Variable Length of Representations","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Lingjun Mao, Rodolfo Corona, Wenhao Yan, Xin Liang, Zineng Tang","submitted_at":"2025-06-04T07:40:33Z","abstract_excerpt":"Most existing vision encoders map images into a fixed-length sequence of tokens, overlooking the fact that different images contain varying amounts of information. For example, a visually complex image (e.g., a cluttered room) inherently carries more information and thus deserves more tokens than a simple image (e.g., a blank wall). To address this inefficiency, we propose DOVE, a dynamic vision encoder that produces a variable number of visual tokens (i.e., continuous representation vectors) to reconstruct each image. Our results show that DOVE significantly reduces the average number of toke"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.03643","kind":"arxiv","version":2},"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/2506.03643/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":"2506.03643","created_at":"2026-07-05T11:16:36.177452+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.03643v2","created_at":"2026-07-05T11:16:36.177452+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.03643","created_at":"2026-07-05T11:16:36.177452+00:00"},{"alias_kind":"pith_short_12","alias_value":"ZXO4C6FBFJSD","created_at":"2026-07-05T11:16:36.177452+00:00"},{"alias_kind":"pith_short_16","alias_value":"ZXO4C6FBFJSD3BVN","created_at":"2026-07-05T11:16:36.177452+00:00"},{"alias_kind":"pith_short_8","alias_value":"ZXO4C6FB","created_at":"2026-07-05T11:16:36.177452+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.04461","citing_title":"ChannelTok: Efficient Flexible-Length Vision Tokenization","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2603.06351","citing_title":"DC-DiT: Adaptive Compute and Elastic Inference for Visual Generation via Dynamic Chunking","ref_index":17,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ZXO4C6FBFJSD3BVNIBU3AMAPWS","json":"https://pith.science/pith/ZXO4C6FBFJSD3BVNIBU3AMAPWS.json","graph_json":"https://pith.science/api/pith-number/ZXO4C6FBFJSD3BVNIBU3AMAPWS/graph.json","events_json":"https://pith.science/api/pith-number/ZXO4C6FBFJSD3BVNIBU3AMAPWS/events.json","paper":"https://pith.science/paper/ZXO4C6FB"},"agent_actions":{"view_html":"https://pith.science/pith/ZXO4C6FBFJSD3BVNIBU3AMAPWS","download_json":"https://pith.science/pith/ZXO4C6FBFJSD3BVNIBU3AMAPWS.json","view_paper":"https://pith.science/paper/ZXO4C6FB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.03643&json=true","fetch_graph":"https://pith.science/api/pith-number/ZXO4C6FBFJSD3BVNIBU3AMAPWS/graph.json","fetch_events":"https://pith.science/api/pith-number/ZXO4C6FBFJSD3BVNIBU3AMAPWS/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ZXO4C6FBFJSD3BVNIBU3AMAPWS/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ZXO4C6FBFJSD3BVNIBU3AMAPWS/action/storage_attestation","attest_author":"https://pith.science/pith/ZXO4C6FBFJSD3BVNIBU3AMAPWS/action/author_attestation","sign_citation":"https://pith.science/pith/ZXO4C6FBFJSD3BVNIBU3AMAPWS/action/citation_signature","submit_replication":"https://pith.science/pith/ZXO4C6FBFJSD3BVNIBU3AMAPWS/action/replication_record"}},"created_at":"2026-07-05T11:16:36.177452+00:00","updated_at":"2026-07-05T11:16:36.177452+00:00"}