{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:IW7BHYYVBVWWQHQETM7PCBRPYM","short_pith_number":"pith:IW7BHYYV","schema_version":"1.0","canonical_sha256":"45be13e3150d6d681e049b3ef1062fc3394d161e0aa5a9151254ad68ea0218bb","source":{"kind":"arxiv","id":"2509.05614","version":3},"attestation_state":"computed","paper":{"title":"SpecPrune-VLA: Accelerating Vision-Language-Action Models via Action-Aware Self-Speculative Pruning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.RO"],"primary_cat":"cs.CV","authors_text":"Guohao Dai, Hanzhen Wang, Jiaming Xu, Jiayi Pan, Yongkang Zhou, Yong-Lu Li, Yushun Xiang","submitted_at":"2025-09-06T06:22:19Z","abstract_excerpt":"Pruning is a typical acceleration technique for compute-bound models by removing computation on unimportant values. Recently, it has been applied to accelerate Vision-Language-Action (VLA) model inference. However, existing acceleration methods focus on local information from the current action step and ignore the global context, leading to >20% success rate drop and limited speedup in some scenarios. In this paper, we point out spatial-temporal consistency in VLA tasks: input images in consecutive steps exhibit high similarity, and propose the key insight that token selection should combine l"},"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":"2509.05614","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2025-09-06T06:22:19Z","cross_cats_sorted":["cs.AI","cs.RO"],"title_canon_sha256":"8a1866f5b593d82f4577887504f0d5a035ef8fa3aca4ecd33dae16cf2181a62c","abstract_canon_sha256":"714ea10ca002a2b156c5c385aaf07abb3c0b8ed941c1fc2abb6da4bdac2232e9"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-26T01:03:15.279630Z","signature_b64":"o/1ldsgMVEidcMBkotP2yzCnj65DEMPesHEHo2cY3EVjsp9K1UFDm4v6hoL8Bv/Vie8TNkdDMA+wGes9NSqnDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"45be13e3150d6d681e049b3ef1062fc3394d161e0aa5a9151254ad68ea0218bb","last_reissued_at":"2026-05-26T01:03:15.278878Z","signature_status":"signed_v1","first_computed_at":"2026-05-26T01:03:15.278878Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"SpecPrune-VLA: Accelerating Vision-Language-Action Models via Action-Aware Self-Speculative Pruning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.RO"],"primary_cat":"cs.CV","authors_text":"Guohao Dai, Hanzhen Wang, Jiaming Xu, Jiayi Pan, Yongkang Zhou, Yong-Lu Li, Yushun Xiang","submitted_at":"2025-09-06T06:22:19Z","abstract_excerpt":"Pruning is a typical acceleration technique for compute-bound models by removing computation on unimportant values. Recently, it has been applied to accelerate Vision-Language-Action (VLA) model inference. However, existing acceleration methods focus on local information from the current action step and ignore the global context, leading to >20% success rate drop and limited speedup in some scenarios. In this paper, we point out spatial-temporal consistency in VLA tasks: input images in consecutive steps exhibit high similarity, and propose the key insight that token selection should combine l"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2509.05614","kind":"arxiv","version":3},"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/2509.05614/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":"2509.05614","created_at":"2026-05-26T01:03:15.278963+00:00"},{"alias_kind":"arxiv_version","alias_value":"2509.05614v3","created_at":"2026-05-26T01:03:15.278963+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2509.05614","created_at":"2026-05-26T01:03:15.278963+00:00"},{"alias_kind":"pith_short_12","alias_value":"IW7BHYYVBVWW","created_at":"2026-05-26T01:03:15.278963+00:00"},{"alias_kind":"pith_short_16","alias_value":"IW7BHYYVBVWWQHQE","created_at":"2026-05-26T01:03:15.278963+00:00"},{"alias_kind":"pith_short_8","alias_value":"IW7BHYYV","created_at":"2026-05-26T01:03:15.278963+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":7,"internal_anchor_count":7,"sample":[{"citing_arxiv_id":"2509.22244","citing_title":"FlashEdit: Decoupling Speed, Structure, and Semantics for Precise Image Editing","ref_index":39,"is_internal_anchor":true},{"citing_arxiv_id":"2603.19199","citing_title":"FASTER: Rethinking Real-Time Flow VLAs","ref_index":84,"is_internal_anchor":true},{"citing_arxiv_id":"2511.18960","citing_title":"AVA-VLA: Improving Vision-Language-Action models with Active Visual Attention","ref_index":46,"is_internal_anchor":true},{"citing_arxiv_id":"2603.01581","citing_title":"KERV: Kinematic-Rectified Speculative Decoding for Embodied VLA Models","ref_index":26,"is_internal_anchor":true},{"citing_arxiv_id":"2603.19199","citing_title":"FASTER: Rethinking Real-Time Flow VLAs","ref_index":84,"is_internal_anchor":true},{"citing_arxiv_id":"2605.13778","citing_title":"Realtime-VLA FLASH: Speculative Inference Framework for Diffusion-based VLAs","ref_index":25,"is_internal_anchor":true},{"citing_arxiv_id":"2605.02739","citing_title":"Latent Bridge: Feature Delta Prediction for Efficient Dual-System Vision-Language-Action Model Inference","ref_index":9,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/IW7BHYYVBVWWQHQETM7PCBRPYM","json":"https://pith.science/pith/IW7BHYYVBVWWQHQETM7PCBRPYM.json","graph_json":"https://pith.science/api/pith-number/IW7BHYYVBVWWQHQETM7PCBRPYM/graph.json","events_json":"https://pith.science/api/pith-number/IW7BHYYVBVWWQHQETM7PCBRPYM/events.json","paper":"https://pith.science/paper/IW7BHYYV"},"agent_actions":{"view_html":"https://pith.science/pith/IW7BHYYVBVWWQHQETM7PCBRPYM","download_json":"https://pith.science/pith/IW7BHYYVBVWWQHQETM7PCBRPYM.json","view_paper":"https://pith.science/paper/IW7BHYYV","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2509.05614&json=true","fetch_graph":"https://pith.science/api/pith-number/IW7BHYYVBVWWQHQETM7PCBRPYM/graph.json","fetch_events":"https://pith.science/api/pith-number/IW7BHYYVBVWWQHQETM7PCBRPYM/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/IW7BHYYVBVWWQHQETM7PCBRPYM/action/timestamp_anchor","attest_storage":"https://pith.science/pith/IW7BHYYVBVWWQHQETM7PCBRPYM/action/storage_attestation","attest_author":"https://pith.science/pith/IW7BHYYVBVWWQHQETM7PCBRPYM/action/author_attestation","sign_citation":"https://pith.science/pith/IW7BHYYVBVWWQHQETM7PCBRPYM/action/citation_signature","submit_replication":"https://pith.science/pith/IW7BHYYVBVWWQHQETM7PCBRPYM/action/replication_record"}},"created_at":"2026-05-26T01:03:15.278963+00:00","updated_at":"2026-05-26T01:03:15.278963+00:00"}