# VLA-OPD: Bridging Offline SFT and Online RL for Vision-Language-Action Models via On-Policy Distillation

arXiv: https://arxiv.org/abs/2603.26666 · HTML: https://mlsanity.com/paper/2603.26666/

## Key facts

- Cited by 13 papers in the map (data through 14 Oct 2026).
- Placed in Reinforcement learning and post-training, one of 13 directions in Vision-language-action models.
- Builds on 28 of its 34 references that are in the map.
- First posted on arXiv 27 Mar 2026.

## Placed in

- Vision-language-action models / [Reinforcement learning and post-training](https://mlsanity.com/vla/reinforcement-learning-and-post-training/)

## Builds on (most cited first, 28 of 28)

- [RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control](https://mlsanity.com/paper/2307.15818/) (2023)
- [OpenVLA: An Open-Source Vision-Language-Action Model](https://mlsanity.com/paper/2406.09246/) (2024)
- [Octo: An Open-Source Generalist Robot Policy](https://mlsanity.com/paper/2405.12213/) (2024)
- [LIBERO: Benchmarking Knowledge Transfer for Lifelong Robot Learning](https://mlsanity.com/paper/2306.03310/) (2023)
- [GR00T N1: An Open Foundation Model for Generalist Humanoid Robots](https://mlsanity.com/paper/2503.14734/) (2025)
- [Fine-Tuning Vision-Language-Action Models: Optimizing Speed and Success](https://mlsanity.com/paper/2502.19645/) (2025)
- [Open X-Embodiment: Robotic Learning Datasets and RT-X Models](https://mlsanity.com/paper/2310.08864/) (2023)
- [RDT-1B: a Diffusion Foundation Model for Bimanual Manipulation](https://mlsanity.com/paper/2410.07864/) (2024)
- [SFT Memorizes, RL Generalizes: A Comparative Study of Foundation Model Post-training](https://mlsanity.com/paper/2501.17161/) (2025)
- [FAST: Efficient Action Tokenization for Vision-Language-Action Models](https://mlsanity.com/paper/2501.09747/) (2025)
- [RoboTwin 2.0: A Scalable Data Generator and Benchmark with Strong Domain Randomization for Robust Bimanual Robotic Manipulation](https://mlsanity.com/paper/2506.18088/) (2025)
- [π*0.6: a VLA That Learns From Experience](https://mlsanity.com/paper/2511.14759/) (2025)
- [HG-DAgger: Interactive Imitation Learning with Human Experts](https://mlsanity.com/paper/1810.02890/) (2018)
- [VLA-RL: Towards Masterful and General Robotic Manipulation with Scalable Reinforcement Learning](https://mlsanity.com/paper/2505.18719/) (2025)
- [RL's Razor: Why Online Reinforcement Learning Forgets Less](https://mlsanity.com/paper/2509.04259/) (2025)
- [ConRFT: A Reinforced Fine-tuning Method for VLA Models via Consistency Policy](https://mlsanity.com/paper/2502.05450/) (2025)
- [SimpleVLA-RL: Scaling VLA Training via Reinforcement Learning](https://mlsanity.com/paper/2509.09674/) (2025)
- [Interactive Post-Training for Vision-Language-Action Models](https://mlsanity.com/paper/2505.17016/) (2025)
- [NORA: A Small Open-Sourced Generalist Vision Language Action Model for Embodied Tasks](https://mlsanity.com/paper/2504.19854/) (2025)
- [What Can RL Bring to VLA Generalization? An Empirical Study](https://mlsanity.com/paper/2505.19789/) (2025)
- [PD-VLA: Accelerating Vision-Language-Action Model Integrated with Action Chunking via Parallel Decoding](https://mlsanity.com/paper/2503.02310/) (2025)
- [Self-Improving Vision-Language-Action Models with Data Generation via Residual RL](https://mlsanity.com/paper/2511.00091/) (2025)
- [VLA-RFT: Vision-Language-Action Reinforcement Fine-tuning with Verified Rewards in World Simulators](https://mlsanity.com/paper/2510.00406/) (2025)
- [The Path Not Taken: RLVR Provably Learns Off the Principals](https://mlsanity.com/paper/2511.08567/) (2025)
- [Reinforcement Fine-Tuning Naturally Mitigates Forgetting in Continual Post-Training](https://mlsanity.com/paper/2507.05386/) (2025)
- [Towards a Unified Understanding of Robot Manipulation: A Comprehensive Survey](https://mlsanity.com/paper/2510.10903/) (2025)
- [STARE-VLA: Progressive Stage-Aware Reinforcement for Fine-Tuning Vision-Language-Action Models](https://mlsanity.com/paper/2512.05107/) (2025)
- [GKD: A General Knowledge Distillation Framework for Large-scale Pre-trained Language Model](https://mlsanity.com/paper/2306.06629/) (2023)

## Built on by (most cited first, 13 of 13)

- [A Survey of On-Policy Distillation for Large Language Models](https://mlsanity.com/paper/2604.00626/) (2026)
- [Filter, Then Reweight: Rethinking Optimization Granularity in On-Policy Distillation](https://mlsanity.com/paper/2606.02684/) (2026)
- [UniSD: Towards a Unified Self-Distillation Framework for Large Language Models](https://mlsanity.com/paper/2605.06597/) (2026)
- [A Formula-Driven Survey and Research Agenda for On-Policy Distillation](https://mlsanity.com/paper/2606.22793/) (2026)
- [WAM-OPD: On-Policy Distillation for World Action Models](https://mlsanity.com/paper/2608.22364/) (2026)
- [Safe Embodied AI for Long-horizon Tasks: A Cross-layer Analysis of Robotic Manipulation](https://mlsanity.com/paper/2606.05660/) (2026)
- [Teach it to stop, not just to click](https://mlsanity.com/paper/2607.17136/) (2026)
- [Coherent Off-Policy Improvement of Large Behavior Models with Learned Rewards](https://mlsanity.com/paper/2606.02194/) (2026)
- [ROAD-VLA: Robust Online Adaptation via Self-Distillation for Vision-Language-Action Models](https://mlsanity.com/paper/2606.25800/) (2026)
- [WAM-OPD: Sharpening World Action Models via On-Policy Distillation](https://mlsanity.com/paper/2609.34250/) (2026)
- [Understanding Off- vs On-Policy Distillation: A Tale of Distinct Training Objectives](https://mlsanity.com/paper/2609.38666/) (2026)
- [DriftOPD: Sequence-Level Reverse-KL Distillation for One-Step VLA Policies](https://mlsanity.com/paper/2610.00317/) (2026)
- [Beyond In-Distribution Preservation: Recovering Generalization in Quantized VLAs via Vulnerability-Oriented Tuning](https://mlsanity.com/paper/2610.05745/) (2026)

## Abstract (arXiv)

Although pre-trained Vision-Language-Action (VLA) models exhibit impressive generalization in robotic manipulation, post-training remains crucial to ensure reliable performance during deployment. However, standard offline Supervised Fine-Tuning (SFT) suffers from distribution shifts and catastrophic forgetting of pre-trained capabilities, while online Reinforcement Learning (RL) struggles with sparse rewards and poor sample efficiency. In this paper, we propose On-Policy VLA Distillation (VLA-OPD), a framework bridging the efficiency of SFT with the robustness of RL. Instead of relying on sparse environmental rewards, VLA-OPD leverages an expert teacher to provide dense, token-level supervision on the student's self-generated trajectories. This enables active error correction on policy-induced states while preserving pre-trained general capabilities through gentle alignment. Crucially, we formulate VLA-OPD via a Reverse-KL objective. Unlike standard Forward-KL that induces mode-covering entropy explosion, or Hard-CE that causes premature entropy collapse, our bounded mode-seeking objective ensures stable policy learning by filtering out the teacher's epistemic uncertainty while maintaining action diversity. Experiments on LIBERO and RoboTwin2.0 benchmarks demonstrate that VLA-OPD significantly improves sample efficiency over RL and robustness over SFT, while effectively mitigating catastrophic forgetting during post-training.