{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:6SJIEY35VDZHUFIUDT5N4FSZED","short_pith_number":"pith:6SJIEY35","schema_version":"1.0","canonical_sha256":"f49282637da8f27a15141cfade165920e8fb880d5482547590038537bcf37203","source":{"kind":"arxiv","id":"2603.05950","version":2},"attestation_state":"computed","paper":{"title":"Energy-Driven Adaptive Visual Token Pruning for Efficient Vision-Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CV","authors_text":"Huangxun Chen, Jialuo He","submitted_at":"2026-03-06T06:32:57Z","abstract_excerpt":"Visual token reduction is critical for accelerating Vision-Language Models (VLMs), since visual inputs are represented as token sequences that introduce substantial computational overhead in the LLM backbone. However, most pruning pipelines treat efficiency primarily as a token selection problem and retain a fixed visual token budget across inputs, overlooking the substantial variation in image information density. We propose E-AdaPrune, an energy driven adaptive pruning framework that determines an image specific token budget from the singular value spectrum of the visual feature matrix and p"},"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":"2603.05950","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-03-06T06:32:57Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"8bec5fdbf45ff389578c86896d62444418d55d5ad2c34feab0a72df13b6d855d","abstract_canon_sha256":"f260b095e9b9be8296939e2cbb9f6c20c29f06853d45ad9538c14c3f2e880130"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f49282637da8f27a15141cfade165920e8fb880d5482547590038537bcf37203","last_reissued_at":"2026-07-31T01:33:20.453714Z","signature_status":"unsigned_v0","first_computed_at":"2026-07-31T01:33:20.453714Z"},"graph_snapshot":{"paper":{"title":"Energy-Driven Adaptive Visual Token Pruning for Efficient Vision-Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CV","authors_text":"Huangxun Chen, Jialuo He","submitted_at":"2026-03-06T06:32:57Z","abstract_excerpt":"Visual token reduction is critical for accelerating Vision-Language Models (VLMs), since visual inputs are represented as token sequences that introduce substantial computational overhead in the LLM backbone. However, most pruning pipelines treat efficiency primarily as a token selection problem and retain a fixed visual token budget across inputs, overlooking the substantial variation in image information density. We propose E-AdaPrune, an energy driven adaptive pruning framework that determines an image specific token budget from the singular value spectrum of the visual feature matrix and p"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2603.05950","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/2603.05950/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":"2603.05950","created_at":"2026-07-31T01:33:20.456774+00:00"},{"alias_kind":"arxiv_version","alias_value":"2603.05950v2","created_at":"2026-07-31T01:33:20.456774+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2603.05950","created_at":"2026-07-31T01:33:20.456774+00:00"},{"alias_kind":"pith_short_12","alias_value":"6SJIEY35VDZH","created_at":"2026-07-31T01:33:20.456774+00:00"},{"alias_kind":"pith_short_16","alias_value":"6SJIEY35VDZHUFIU","created_at":"2026-07-31T01:33:20.456774+00:00"},{"alias_kind":"pith_short_8","alias_value":"6SJIEY35","created_at":"2026-07-31T01:33:20.456774+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":3,"sample":[{"citing_arxiv_id":"2605.25343","citing_title":"Toward Native Multimodal Modeling: A Roadmap","ref_index":260,"is_internal_anchor":true},{"citing_arxiv_id":"2605.30126","citing_title":"PARCEL: Pool-Anchored Resampling with Conditioned Elastic Queries for Efficient Vision-Language Understanding","ref_index":44,"is_internal_anchor":true},{"citing_arxiv_id":"2604.24391","citing_title":"FreqCache: Accelerating Embodied VLN Models with Adaptive Frequency-Guided Token Caching","ref_index":10,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/6SJIEY35VDZHUFIUDT5N4FSZED","json":"https://pith.science/pith/6SJIEY35VDZHUFIUDT5N4FSZED.json","graph_json":"https://pith.science/api/pith-number/6SJIEY35VDZHUFIUDT5N4FSZED/graph.json","events_json":"https://pith.science/api/pith-number/6SJIEY35VDZHUFIUDT5N4FSZED/events.json","paper":"https://pith.science/paper/6SJIEY35"},"agent_actions":{"view_html":"https://pith.science/pith/6SJIEY35VDZHUFIUDT5N4FSZED","download_json":"https://pith.science/pith/6SJIEY35VDZHUFIUDT5N4FSZED.json","view_paper":"https://pith.science/paper/6SJIEY35","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2603.05950&json=true","fetch_graph":"https://pith.science/api/pith-number/6SJIEY35VDZHUFIUDT5N4FSZED/graph.json","fetch_events":"https://pith.science/api/pith-number/6SJIEY35VDZHUFIUDT5N4FSZED/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/6SJIEY35VDZHUFIUDT5N4FSZED/action/timestamp_anchor","attest_storage":"https://pith.science/pith/6SJIEY35VDZHUFIUDT5N4FSZED/action/storage_attestation","attest_author":"https://pith.science/pith/6SJIEY35VDZHUFIUDT5N4FSZED/action/author_attestation","sign_citation":"https://pith.science/pith/6SJIEY35VDZHUFIUDT5N4FSZED/action/citation_signature","submit_replication":"https://pith.science/pith/6SJIEY35VDZHUFIUDT5N4FSZED/action/replication_record"}},"created_at":"2026-07-31T01:33:20.456774+00:00","updated_at":"2026-07-31T01:33:20.456774+00:00"}