{"paper":{"title":"FASTER: Rethinking Real-Time Flow VLAs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"A Horizon-Aware Schedule lets flow-based VLAs complete the first action's denoising in one step instead of many while keeping the full trajectory intact.","cross_cats":["cs.CV"],"primary_cat":"cs.RO","authors_text":"Hengshuang Zhao, Jinghua Hou, Junyi Li, Kaixin Ding, Xianzhe Fan, Yuxiang Lu, Zhe Liu, Zhenya Yang","submitted_at":"2026-03-19T17:51:37Z","abstract_excerpt":"Real-time execution is crucial for deploying Vision-Language-Action (VLA) models in the physical world. Existing asynchronous inference methods primarily optimize trajectory smoothness, but neglect the critical latency in reacting to environmental changes. By rethinking the notion of reaction in action chunking policies, this paper presents a systematic analysis of the factors governing reaction time. We show that reaction time follows a uniform distribution determined jointly by the Time to First Action (TTFA) and the execution horizon. Moreover, we reveal that the standard practice of applyi"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"By introducing a Horizon-Aware Schedule, FASTER adaptively prioritizes near-term actions during flow sampling, compressing the denoising of the immediate reaction by tenfold into a single step, while preserving the quality of long-horizon trajectory.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The assumption that adaptively prioritizing near-term denoising steps will preserve long-horizon trajectory quality without introducing artifacts or instability that would appear only under real-world dynamics or longer horizons.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"FASTER uses a horizon-aware flow sampling schedule to compress immediate-action denoising to one step, slashing effective reaction latency in real-robot VLA deployments.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"A Horizon-Aware Schedule lets flow-based VLAs complete the first action's denoising in one step instead of many while keeping the full trajectory intact.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"1d07212ecb226a77b11a95c0d2692a8a740c6293eda718e5ccbeb5ad1dae9fb2"},"source":{"id":"2603.19199","kind":"arxiv","version":3},"verdict":{"id":"9bb9c623-32ff-4583-b18a-41d33c5f38da","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T07:59:15.266339Z","strongest_claim":"By introducing a Horizon-Aware Schedule, FASTER adaptively prioritizes near-term actions during flow sampling, compressing the denoising of the immediate reaction by tenfold into a single step, while preserving the quality of long-horizon trajectory.","one_line_summary":"FASTER uses a horizon-aware flow sampling schedule to compress immediate-action denoising to one step, slashing effective reaction latency in real-robot VLA deployments.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The assumption that adaptively prioritizing near-term denoising steps will preserve long-horizon trajectory quality without introducing artifacts or instability that would appear only under real-world dynamics or longer horizons.","pith_extraction_headline":"A Horizon-Aware Schedule lets flow-based VLAs complete the first action's denoising in one step instead of many while keeping the full trajectory intact."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2603.19199/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":2,"snapshot_sha256":"bb449034aca94f9c76b6c12c3ee35833e451d174e46447b1673486d2e2d7118c"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}