Enhancing Vision-Language Model Training with Reinforcement Learning in Synthetic Worlds for Real-World Success

• URL: http://arxiv.org/abs/2508.04280v1
• Date: Wed, 06 Aug 2025 10:08:48 GMT
• Title: Enhancing Vision-Language Model Training with Reinforcement Learning in Synthetic Worlds for Real-World Success
• Authors: George Bredis, Stanislav Dereka, Viacheslav Sinii, Ruslan Rakhimov, Daniil Gavrilov,
• Abstract summary: We introduce Vision-Language Decoupled Actor-Critic (VLDAC), a lightweight reinforcement learning algorithm.<n>VLDAC applies PPO updates to action while learning value only at the environment-step level.<n>Training a single VLM with VLDAC in one inexpensive simulator at a time produces policies that generalize widely.
• Score: 5.617637951327993
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Interactive multimodal agents must convert raw visual observations into coherent sequences of language-conditioned actions -- a capability that current vision-language models (VLMs) still lack. Earlier reinforcement-learning (RL) efforts could, in principle, endow VLMs with such skills, but they have seldom tested whether the learned behaviours generalize beyond their training simulators, and they depend either on brittle hyperparameter tuning or on dense-reward environments with low state variability. We introduce Vision-Language Decoupled Actor-Critic (VL-DAC), a lightweight, hyperparameter-free RL algorithm. VL-DAC applies PPO updates to action tokens while learning value only at the environment-step level: an arrangement, to our knowledge, not previously explored for large VLMs or LLMs. This simple decoupling removes unstable weighting terms and yields faster, more reliable convergence. Training a single VLM with VL-DAC in one inexpensive simulator at a time (MiniWorld, Gym-Cards, ALFWorld, or WebShop) already produces policies that generalize widely: +50\% relative on BALROG (game-centric agentic control), +5\% relative on the hardest part of VSI-Bench (spatial planning), and +2\% on VisualWebBench (web navigation), all without degrading general image understanding accuracy. These results provide the first evidence that a simple RL algorithm can train VLMs entirely in cheap synthetic worlds while delivering measurable gains on real-image agentic, spatial-reasoning, and web-navigation benchmarks.

Related papers

• SmolVLA: A Vision-Language-Action Model for Affordable and Efficient Robotics [37.82471658370026]
  Vision-language models (VLMs) pretrained on large-scale multimodal datasets encode rich visual and linguistic knowledge.<n>We present SmolVLA, a small, efficient, and community-driven VLA that drastically reduces both training and inference costs.
  arXiv  Detail & Related papers  (2025-06-02T16:30:19Z)
• ViaRL: Adaptive Temporal Grounding via Visual Iterated Amplification Reinforcement Learning [68.76048244253582]
  We introduce ViaRL, the first framework to leverage rule-based reinforcement learning (RL) for optimizing frame selection in video understanding.<n>ViaRL utilizes the answer accuracy of a downstream model as a reward signal to train a frame selector through trial-and-error.<n>ViaRL consistently delivers superior temporal grounding performance and robust generalization across diverse video understanding tasks.
  arXiv  Detail & Related papers  (2025-05-21T12:29:40Z)
• VLM-RL: A Unified Vision Language Models and Reinforcement Learning Framework for Safe Autonomous Driving [1.3107174618549584]
  Reinforcement learning (RL)-based methods for learning driving policies have gained increasing attention in the autonomous driving community.<n>Traditional RL approaches rely on manually engineered rewards, which require extensive human effort and often lack generalizability.<n>We propose textbfVLM-RL, a unified framework that integrates pre-trained Vision-Language Models (VLMs) with RL to generate reward signals.
  arXiv  Detail & Related papers  (2024-12-20T04:08:11Z)
• Vision Language Models are In-Context Value Learners [89.29486557646624]
  We present Generative Value Learning (GVL), a universal value function estimator that leverages the world knowledge embedded in vision-language models (VLMs) to predict task progress. Without any robot or task specific training, GVL can in-context zero-shot and few-shot predict effective values for more than 300 distinct real-world tasks.
  arXiv  Detail & Related papers  (2024-11-07T09:17:50Z)
• Mono-InternVL: Pushing the Boundaries of Monolithic Multimodal Large Language Models with Endogenous Visual Pre-training [48.455597568212944]
  We present Mono-InternVL, a novel monolithic MLLM that seamlessly integrates a set of visual experts via a multimodal mixture-of-experts structure.<n>In particular, EViP is designed as a progressive learning process for visual experts, which aims to fully exploit the visual knowledge from noisy data to high-quality data.
  arXiv  Detail & Related papers  (2024-10-10T17:59:22Z)
• How Can LLM Guide RL? A Value-Based Approach [68.55316627400683]
  Reinforcement learning (RL) has become the de facto standard practice for sequential decision-making problems by improving future acting policies with feedback. Recent developments in large language models (LLMs) have showcased impressive capabilities in language understanding and generation, yet they fall short in exploration and self-improvement capabilities. We develop an algorithm named LINVIT that incorporates LLM guidance as a regularization factor in value-based RL, leading to significant reductions in the amount of data needed for learning.
  arXiv  Detail & Related papers  (2024-02-25T20:07:13Z)
• Test-Time Adaptation with CLIP Reward for Zero-Shot Generalization in Vision-Language Models [76.410400238974]
  We propose TTA with feedback to rectify the model output and prevent the model from becoming blindly confident. A CLIP model is adopted as the reward model during TTA and provides feedback for the VLM. The proposed textitreinforcement learning with CLIP feedback(RLCF) framework is highly flexible and universal.
  arXiv  Detail & Related papers  (2023-05-29T11:03:59Z)
• Cheap and Quick: Efficient Vision-Language Instruction Tuning for Large Language Models [77.2078051555533]
  We propose a novel and affordable solution for the effective VL adaption of large language models (LLMs) Instead of using large neural networks to connect the image encoder and LLM, MMA adopts lightweight modules, i.e., adapters. MMA is also equipped with a routing algorithm to help LLMs achieve an automatic shift between single- and multi-modal instructions.
  arXiv  Detail & Related papers  (2023-05-24T11:06:15Z)

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.