Beyond Imitation: Reinforcement Learning-Based Sim-Real Co-Training for VLA Models

• URL: http://arxiv.org/abs/2602.12628v3
• Date: Fri, 06 Mar 2026 08:46:11 GMT
• Title: Beyond Imitation: Reinforcement Learning-Based Sim-Real Co-Training for VLA Models
• Abstract summary: We propose an underlinetextitRL-based sim-real underlinetextitCo-training modify(RL-Co) framework.<n> RL co-training yields stronger generalization to unseen task variations and substantially improved real-world data efficiency.
• Score: 35.41613201907666
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Simulation offers a scalable and low-cost way to enrich vision-language-action (VLA) training, reducing reliance on expensive real-robot demonstrations. However, most sim-real co-training methods rely on supervised fine-tuning (SFT), which treats simulation as a static source of demonstrations and does not exploit large-scale closed-loop interaction. Consequently, real-world gains and generalization are often limited. In this paper, we propose an \underline{\textit{RL}}-based sim-real \underline{\textit{Co}}-training \modify{(RL-Co)} framework that leverages interactive simulation while preserving real-world capabilities. Our method follows a generic two-stage design: we first warm-start the policy with SFT on a mixture of real and simulated demonstrations, then fine-tune it with reinforcement learning in simulation while adding an auxiliary supervised loss on real-world data to anchor the policy and mitigate catastrophic forgetting. We evaluate our framework on four real-world tabletop manipulation tasks using two representative VLA architectures, OpenVLA and $π_{0.5}$, and observe consistent improvements over real-only fine-tuning and SFT-based co-training, including +24% real-world success on OpenVLA and +20% on $π_{0.5}$. Beyond higher success rates, RL co-training yields stronger generalization to unseen task variations and substantially improved real-world data efficiency, providing a practical and scalable pathway for leveraging simulation to enhance real-robot deployment.

Related papers

• Object-Centric Residual RL for Zero-Shot Sim-to-Real VLA Enhancement [52.93267443851685]
  Vision-Language-Action (VLA) models can generalize across diverse manipulation tasks.<n>Can a reinforcement learning policy trained purely in simulation improve the robustness of real-world VLAs zero-shot?<n>We propose an object-centric residual RL framework that refines VLA actions using object poses.
  arXiv  Detail & Related papers  (2026-06-17T11:36:54Z)
• RLFTSim: Realistic and Controllable Multi-Agent Traffic Simulation via Reinforcement Learning Fine-Tuning [19.985714275376726]
  We introduce RLFTSim, a reinforcement-learning-based fine-tuning framework.<n>It enhances scenario realism by aligning simulator rollouts with real-world data distributions.<n>We also demonstrate the effectiveness of our approach for distilling traffic simulation controllability through goal conditioning.
  arXiv  Detail & Related papers  (2026-05-18T18:57:13Z)
• ComSim: Building Scalable Real-World Robot Data Generation via Compositional Simulation [90.4702774169675]
  We propose a novel hybrid approach called Compositional Simulation, which combines classical simulation and neural simulation.<n>Our approach utilizes a closed-loop real-sim-real data augmentation pipeline, leveraging a small amount of real-world data.<n>We train a neural simulator to transform classical simulation videos into real-world representations, improving the accuracy of policy models trained in real-world environments.
  arXiv  Detail & Related papers  (2026-04-13T12:25:45Z)
• DexWorldModel: Causal Latent World Modeling towards Automated Learning of Embodied Tasks [54.32016216994156]
  We introduce the Causal Latent World Model (CLWM), which employs DINOv3 features as generative targets to disentangle interaction semantics from visual noise.<n>To overcome memory scaling, CLWM features a Dual-State Test-Time Training (TTT) Memory that guarantees a strict $mathcalO(1)$ footprint for long-horizon tasks.<n>We present EmbodiChain, an online framework that establishes the Efficiency Law by injecting an infinite flow of physics-grounded trajectories during training.
  arXiv  Detail & Related papers  (2026-04-13T03:19:36Z)
• Reinforcement World Model Learning for LLM-based Agents [60.65003139516272]
  Reinforcement World Model Learning (RWML) is a self-conditioned method that learns action-supervised world models for LLM-based agents.<n>Our method aligns simulated next states produced by the model with realized next states observed from the environment.<n>We evaluate our method on ALFWorld and $2$ Bench and observe significant gains over the base model, despite being entirely self-supervised.
  arXiv  Detail & Related papers  (2026-02-05T16:30:08Z)
• Scaling Agent Learning via Experience Synthesis [100.42712232390532]
  Reinforcement learning can empower autonomous agents by enabling self-improvement through interaction.<n>But its practical adoption remains challenging due to costly rollouts, limited task diversity, unreliable reward signals, and infrastructure complexity.<n>We introduce DreamGym, the first unified framework designed to synthesize diverse experiences with scalability in mind.
  arXiv  Detail & Related papers  (2025-11-05T18:58:48Z)
• Simulating Environments with Reasoning Models for Agent Training [55.98861707136674]
  Building bespoke environments for training is heavy, brittle, and limits progress.<n>We propose two frameworks: Simia-SFT and Simia-RL.<n>Simia-SFT and Simia-RL enable scalable agent training without environment engineering.
  arXiv  Detail & Related papers  (2025-11-03T18:29:57Z)
• VLA-RFT: Vision-Language-Action Reinforcement Fine-tuning with Verified Rewards in World Simulators [38.880852900641]
  Vision-Language-Action (VLA) models enable embodied decision-making but rely heavily on imitation learning.<n>We introduce VLA-RFT, a reinforcement fine-tuning framework that leverages a data-driven world model as a controllable simulator.<n>With fewer than 400 fine-tuning steps, VLA-RFT surpasses strong supervised baselines and achieves greater efficiency than simulator-based RL.
  arXiv  Detail & Related papers  (2025-10-01T01:33:10Z)
• Generalizable Domain Adaptation for Sim-and-Real Policy Co-Training [21.855770200309674]
  We propose a unified sim-and-real co-training framework for learning generalizable manipulation policies.<n>We show it can leverage abundant simulation data to achieve up to a 30% improvement in the real-world success rate.
  arXiv  Detail & Related papers  (2025-09-23T04:32:53Z)
• Overcoming the Sim-to-Real Gap: Leveraging Simulation to Learn to Explore for Real-World RL [25.991354823569033]
  We show that in many regimes, while direct sim2real transfer may fail, we can utilize the simulator to learn a set of emphexploratory policies. In particular, in the setting of low-rank MDPs, we show that coupling these exploratory policies with simple, practical approaches. This is the first evidence that simulation transfer yields a provable gain in reinforcement learning in settings where direct sim2real transfer fails.
  arXiv  Detail & Related papers  (2024-10-26T19:12:27Z)
• A Conservative Approach for Few-Shot Transfer in Off-Dynamics Reinforcement Learning [3.1515473193934778]
  Off-dynamics Reinforcement Learning seeks to transfer a policy from a source environment to a target environment characterized by distinct yet similar dynamics. We propose an innovative approach inspired by recent advancements in Imitation Learning and conservative RL algorithms.
  arXiv  Detail & Related papers  (2023-12-24T13:09:08Z)
• TrafficSim: Learning to Simulate Realistic Multi-Agent Behaviors [74.67698916175614]
  We propose TrafficSim, a multi-agent behavior model for realistic traffic simulation. In particular, we leverage an implicit latent variable model to parameterize a joint actor policy. We show TrafficSim generates significantly more realistic and diverse traffic scenarios as compared to a diverse set of baselines.
  arXiv  Detail & Related papers  (2021-01-17T00:29:30Z)
• RL-CycleGAN: Reinforcement Learning Aware Simulation-To-Real [74.45688231140689]
  We introduce the RL-scene consistency loss for image translation, which ensures that the translation operation is invariant with respect to the Q-values associated with the image. We obtain RL-CycleGAN, a new approach for simulation-to-real-world transfer for reinforcement learning.
  arXiv  Detail & Related papers  (2020-06-16T08:58:07Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.