{"paper":{"title":"Dynamic Execution Commitment of Vision-Language-Action Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Vision-language-action models can adaptively commit to action sequences by verifying the longest consistent prefix through consensus sampling and invariance checks.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Boying Li, Feng Chen, Xianghui Wang, Yefei He, Yicheng Wu, Yuxuan Chen, Zeyu Zhang","submitted_at":"2026-05-12T05:52:58Z","abstract_excerpt":"Vision-Language-Action (VLA) models predominantly adopt action chunking, i.e., predicting and committing to a short horizon of consecutive low-level actions in a single forward pass, to amortize the inference cost of large-scale backbones and reduce per-step latency. However, committing these multi-step predictions to real-world execution requires balancing success rate against inference efficiency, a decision typically governed by fixed execution horizons tuned per task. Such heuristics ignore the state-dependent nature of predictive reliability, leading to brittle performance in dynamic or o"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Consequently, the execution horizon emerges as the longest verifiable prefix satisfying both internal model logic and sequential execution constraints.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That a trajectory-wise consensus score computed via group sampling, combined with the two proposed verification rules, reliably identifies state-dependent predictive reliability in dynamic or out-of-distribution settings.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"A3 determines the execution horizon in VLA models as the longest prefix of actions that passes consensus-based verification and sequential consistency checks.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Vision-language-action models can adaptively commit to action sequences by verifying the longest consistent prefix through consensus sampling and invariance checks.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"4f8b4fc661d6c72e804cb8f28a49f48685fa365d46e2f2701c117076215d6a88"},"source":{"id":"2605.11567","kind":"arxiv","version":2},"verdict":{"id":"cf67dc37-556f-4b28-b58e-c7e02c22bfe6","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-13T01:41:53.869503Z","strongest_claim":"Consequently, the execution horizon emerges as the longest verifiable prefix satisfying both internal model logic and sequential execution constraints.","one_line_summary":"A3 determines the execution horizon in VLA models as the longest prefix of actions that passes consensus-based verification and sequential consistency checks.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That a trajectory-wise consensus score computed via group sampling, combined with the two proposed verification rules, reliably identifies state-dependent predictive reliability in dynamic or out-of-distribution settings.","pith_extraction_headline":"Vision-language-action models can adaptively commit to action sequences by verifying the longest consistent prefix through consensus sampling and invariance checks."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.11567/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-19T12:33:46.736920Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_title_agreement","ran_at":"2026-05-19T09:31:18.516352Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T08:19:30.518076Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"6b68a89a42c9ee3502a6e4e9a815f00c1c75f266a8f4a86c0851f2705baf3c45"},"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"}