{"paper":{"title":"mimic-video: Video-Action Models for Generalizable Robot Control Beyond VLAs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Pretrained video models plus a flow-matching decoder let robots learn manipulation with far less data than vision-language-action models.","cross_cats":["cs.AI","cs.CV","cs.LG"],"primary_cat":"cs.RO","authors_text":"Benedek Forrai, Elvis Nava, Jonas Pai, Liam Achenbach, Oier Mees, Victoriano Montesinos","submitted_at":"2025-12-17T18:47:31Z","abstract_excerpt":"Prevailing Vision-Language-Action Models (VLAs) for robotic manipulation are built upon vision-language backbones pretrained on large-scale, but disconnected static web data. As a result, despite improved semantic generalization, the policy must implicitly infer complex physical dynamics and temporal dependencies solely from robot trajectories. This reliance creates an unsustainable data burden, necessitating continuous, large-scale expert data collection to compensate for the lack of innate physical understanding. We contend that while vision-language pretraining effectively captures semantic"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Our extensive evaluation shows that our approach achieves state-of-the-art performance on simulated and real-world robotic manipulation tasks, improving sample efficiency by 10x and convergence speed by 2x compared to traditional VLA architectures.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That a pretrained internet video model already captures sufficient physical causality and temporal dynamics so that the remaining task reduces cleanly to low-level control via the flow-matching decoder.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"mimic-video combines internet video pretraining with a flow-matching decoder to achieve state-of-the-art robotic manipulation performance with 10x better sample efficiency than vision-language-action models.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Pretrained video models plus a flow-matching decoder let robots learn manipulation with far less data than vision-language-action models.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"91dedd1877388b6d03a9e70fdf923b5d2854898cf737ce91e32b10ed4fa5f0b8"},"source":{"id":"2512.15692","kind":"arxiv","version":2},"verdict":{"id":"7803f969-855a-4ac9-8d9b-e11480a9afb5","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T10:35:14.719280Z","strongest_claim":"Our extensive evaluation shows that our approach achieves state-of-the-art performance on simulated and real-world robotic manipulation tasks, improving sample efficiency by 10x and convergence speed by 2x compared to traditional VLA architectures.","one_line_summary":"mimic-video combines internet video pretraining with a flow-matching decoder to achieve state-of-the-art robotic manipulation performance with 10x better sample efficiency than vision-language-action models.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That a pretrained internet video model already captures sufficient physical causality and temporal dynamics so that the remaining task reduces cleanly to low-level control via the flow-matching decoder.","pith_extraction_headline":"Pretrained video models plus a flow-matching decoder let robots learn manipulation with far less data than vision-language-action models."},"references":{"count":64,"sample":[{"doi":"","year":2025,"title":"V-JEPA 2: Self-Supervised Video Models Enable Understanding, Prediction and Planning","work_id":"a9c28401-f16a-4933-89f0-788e2f94e52b","ref_index":1,"cited_arxiv_id":"2506.09985","is_internal_anchor":true},{"doi":"","year":2024,"title":"PaliGemma: A versatile 3B VLM for transfer","work_id":"df6f48b3-5792-47c7-9614-cb856ea31ad9","ref_index":2,"cited_arxiv_id":"2407.07726","is_internal_anchor":true},{"doi":"","year":2024,"title":"$\\pi_0$: A Vision-Language-Action Flow Model for General Robot Control","work_id":"f790abdc-a796-482f-a40d-f8ee035ecfc2","ref_index":3,"cited_arxiv_id":"2410.24164","is_internal_anchor":true},{"doi":"","year":2023,"title":"RoboCat : A self-improving foundation agent for robotic manipulation","work_id":"143e7731-0488-4088-8e23-63f9d4140118","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2021,"title":"Emerging Properties in Self-Supervised Vision Transformers","work_id":"6b124bd1-c9f1-4251-96c1-2683f7f17a64","ref_index":5,"cited_arxiv_id":"2104.14294","is_internal_anchor":true}],"resolved_work":64,"snapshot_sha256":"e667f1ad075990562f559a60638d466104c24301e241d90a6aba90e8135b6df8","internal_anchors":35},"formal_canon":{"evidence_count":2,"snapshot_sha256":"48ee4a283a4e41a2e0a0249d42b1a3eab7bd9a69359fe66ac8c678785db4eb0b"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}