Metraipolis › archive › dc_d6010fa843cd
RAFT: Reasoning-Aligned Fine-Tuning Improves Robustness of VLA Post-Training — arXiv:2611.04368
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2611.04368
- License (as archived)
- CC0-1.0
- Archived
- 2026-11-04T19:05:03.285Z
- SHA-256 of the archived text
113f6742d8dab2b361a7c9e1521bd4da382c486a9924d1c0bc6ac51cc4ce65bb- Stable URL of this page
https://metraipolis.com/documento/dc_d6010fa843cd
How to cite
RAFT: Reasoning-Aligned Fine-Tuning Improves Robustness of VLA Post-Training — arXiv:2611.04368 — arXiv (en), https://arxiv.org/abs/2611.04368 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_d6010fa843cd (Metraipolis archive dc_d6010fa843cd, sha256 113f6742d8dab2b361a7c9e1521bd4da382c486a9924d1c0bc6ac51cc4ce65bb)
BibTeX
@misc{dcd6010fa843cd,
title = {RAFT: Reasoning-Aligned Fine-Tuning Improves Robustness of VLA Post-Training — arXiv:2611.04368},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2611.04368},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_d6010fa843cd; sha256 113f6742d8dab2b361a7c9e1521bd4da382c486a9924d1c0bc6ac51cc4ce65bb},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.04368. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Adapting a generalist vision-language-action (VLA) policy to a target task with limited demonstrations can improve nominal task success while reducing robustness to variations in appearance, objects, and instructions. We introduce reasoning-aligned fine-tuning (RAFT), a two-stage post-training procedure: RAFT first trains low-rank adapters on task-relevant reasoning annotations, then freezes them while fine-tuning the policy on action demonstrations. On a physical DROID robot, RAFT achieves $75.0\%$ out-of-distribution success, compared with $27.9\%$ for plain fine-tuning; on a multi-stage bimanual YAM task, it reaches $70.0\%$ mean success, compared with $36.0\%$ for the strongest baseline. RAFT also improves success under visual, object, and instruction shifts in simulated pick-and-place and open-drawer tasks. Compared with plain fine-tuning, RAFT policies place more attention on annotated task regions and yield higher linear-probe accuracy for object identity in later layers. A broader reasoning-data mixture further improves downstream success across demonstration budgets. Tema degli esploratori: statistica e scienza dei dati (elenco arXiv cs.CV, nuovi lavori del giorno). Categorie arXiv: Robotics (cs.RO) ; Computer Vision and Pattern Recognition (cs.CV). Pagina del lavoro: https://arxiv.org/abs/2611.04368 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-11-04T18:39:37.601Z dalla pagina https://arxiv.org/list/cs.CV/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).