RAFT: Reasoning-Aligned Fine-Tuning Improves Robustness of VLA Post-Training
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
- PaLM 2 Technical Report
- GR00T N1: An Open Foundation Model for Generalist Humanoid Robots
- $\pi_0$: A Vision-Language-Action Flow Model for General Robot Control
- RoboCat: A Self-Improving Generalist Agent for Robotic Manipulation
- Steerable Vision-Language-Action Policies for Embodied Reasoning and Hierarchical Control
- Open X-Embodiment: Robotic Learning Datasets and RT-X Models
- MolmoB0T: Large-Scale Simulation Enables Zero-Shot Manipulation
- MolmoAct2: Action Reasoning Models for Real-world Deployment
- Enhancing Generalization in Vision-Language-Action Models by Preserving Pretrained Representations
- Breaking Lock-In: Preserving Steerability under Low-Data VLA Post-Training
- Don't Blind Your VLA: Aligning Visual Representations for OOD Generalization
- Policy Consolidation for Continual Reinforcement Learning
- DROID: A Large-Scale In-The-Wild Robot Manipulation Dataset
- OpenVLA: An Open-Source Vision-Language-Action Model
- Fine-Tuning Vision-Language-Action Models: Optimizing Speed and Success
- MolmoSpaces: A Large-Scale Open Ecosystem for Robot Navigation and Manipulation
- ControlVLA: Few-shot Object-centric Adaptation for Pre-trained Vision-Language-Action Models
- VLS: Steering Pretrained Robot Policies via Vision-Language Models
- SOP: A Scalable Online Post-Training System for Vision-Language-Action Models
- FAST: Efficient Action Tokenization for Vision-Language-Action Models
Related papers
- Rethinking Actions: Continuity-Aware Priors for World-Model Planning
- MOSAIC: Modular Obstacle and Spatial-Aware Framework for Manipulation In Cluttered Scenes
- SAGAS: Semantic-Aware Graph-Assisted Stitching for Offline Temporal Logic Planning
- DexSeed: Expanding Human Hand-Object Interactions into Robot Demonstrations for Dexterous Policy Learning
- LaST0.5: A Generalist World Action Model with Interaction-Centric Latent Reasoning
- CIALLO: Learning Shared Intent Representations for Cross-Embodiment Manipulation