RAFT: Reasoning-Aligned Fine-Tuning Improves Robustness of VLA Post-Training

arXiv:2611.04368 · cs.RO, cs.CV · Submitted 2026-11-03 · Read on arXiv

cs.RO, cs.CV

Submitted: 2026-11-03

Updated: 2026-11-03

Code: https://github.com/princeton-prism/RAFT

License: http://creativecommons.org/licenses/by/4.0/

The gist: 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…

Terminology

Abstract

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.

Sources

Related papers