Vision-Zero: Scalable VLM Self-Improvement via Strategic Gamified Self-Play

• URL: http://arxiv.org/abs/2509.25541v1
• Date: Mon, 29 Sep 2025 21:55:55 GMT
• Title: Vision-Zero: Scalable VLM Self-Improvement via Strategic Gamified Self-Play
• Abstract summary: Vision-Zero is a framework enabling vision-language models (VLMs) self-improvement through competitive visual games generated from arbitrary image pairs.<n> Vision-Zero trains VLMs in "Who Is the Spy"-style games, where the models engage in strategic reasoning and actions across multiple roles.<n>Unlike existing gamified frameworks, Vision-Zero can generate games from arbitrary images, thereby enhancing the model's reasoning ability across diverse domains.
• Score: 27.15191616995172
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Although reinforcement learning (RL) can effectively enhance the reasoning capabilities of vision-language models (VLMs), current methods remain heavily dependent on labor-intensive datasets that require extensive manual construction and verification, leading to extremely high training costs and consequently constraining the practical deployment of VLMs. To address this challenge, we propose Vision-Zero, a domain-agnostic framework enabling VLM self-improvement through competitive visual games generated from arbitrary image pairs. Specifically, Vision-Zero encompasses three main attributes: (1) Strategic Self-Play Framework: Vision-Zero trains VLMs in "Who Is the Spy"-style games, where the models engage in strategic reasoning and actions across multiple roles. Through interactive gameplay, models autonomously generate their training data without human annotation. (2) Gameplay from Arbitrary Images: Unlike existing gamified frameworks, Vision-Zero can generate games from arbitrary images, thereby enhancing the model's reasoning ability across diverse domains and showing strong generalization to different tasks. We demonstrate this versatility using three distinct types of image datasets: CLEVR-based synthetic scenes, charts, and real-world images. (3) Sustainable Performance Gain: We introduce Iterative Self-Play Policy Optimization (Iterative-SPO), a novel training algorithm that alternates between Self-Play and reinforcement learning with verifiable rewards (RLVR), mitigating the performance plateau often seen in self-play-only training and achieving sustained long-term improvements. Despite using label-free data, Vision-Zero achieves state-of-the-art performance on reasoning, chart question answering, and vision-centric understanding tasks, surpassing other annotation-based methods. Models and code has been released at https://github.com/wangqinsi1/Vision-Zero.

Related papers

• From Human Videos to Robot Manipulation: A Survey on Scalable Vision-Language-Action Learning with Human-Centric Data [71.22409934108924]
  Human videos are abundant and capture rich interactions, providing diverse semantic and physical cues for real-world manipulation.<n>This survey provides a unified view of how human videos are transformed into effective knowledge for Vision-Language-Action (VLA) models.<n>We highlight three key open challenges in this area: structuring unstructured videos into training-ready episodes, grounding video-derived supervision into robot-executable actions under embodiment and viewpoint heterogeneity, and designing evaluation protocols that better predict real-world deployment performance and transfer efficiency.
  arXiv  Detail & Related papers  (2026-05-18T06:19:16Z)
• Through the Lens of Contrast: Self-Improving Visual Reasoning in VLMs [60.93949629734977]
  We propose Visual Contrastive Self-Taught Reasoner (VC-STaR) to mitigate hallucinations in model-generated rationales.<n>We collect a diverse suite of VQA datasets, curate contrastive pairs according to multi-modal similarity, and generate rationales using VC-STaR.<n>Extensive experiments show that VC-STaR not only outperforms existing self-improving approaches but also surpasses models finetuned on the SoTA visual reasoning datasets.
  arXiv  Detail & Related papers  (2026-03-03T03:18:31Z)
• Scaling Agentic Reinforcement Learning for Tool-Integrated Reasoning in VLMs [76.47326680870783]
  We introduce VISTA-Gym, a training environment for incentivizing tool-integrated visual reasoning capabilities in vision-language models (VLMs)<n> VISTA-Gym unifies diverse real-world multimodal reasoning tasks with a standardized interface for visual tools.<n>We show that VISTA-R1-8B outperforms state-of-the-art baselines with similar sizes by 9.51%-18.72% in 11 public reasoning-intensive VQA benchmarks.
  arXiv  Detail & Related papers  (2025-11-24T22:58:26Z)
• VisPlay: Self-Evolving Vision-Language Models from Images [16.43670977857598]
  Reinforcement learning (RL) provides a principled framework for improving Vision-Language Models (VLMs) on complex tasks.<n>We introduce VisPlay, a self-evolving RL framework that enablesVLMs to autonomously improve their reasoning abilities using large amounts of unlabeled image data.
  arXiv  Detail & Related papers  (2025-11-19T17:55:15Z)
• Towards Self-Refinement of Vision-Language Models with Triangular Consistency [16.24217978112331]
  Vision-Language Models (VLMs) integrate visual knowledge with the analytical capabilities of Large Language Models (LLMs)<n>This study validates that VLMs possess inherent self-refinement capabilities, enabling them to generate high-quality supervised data without external inputs.
  arXiv  Detail & Related papers  (2025-10-12T07:37:47Z)
• Autoregressive Semantic Visual Reconstruction Helps VLMs Understand Better [44.15671594378141]
  We introduce Autoregressive Semantic Visual Reconstruction (ASVR), which enables joint learning of visual and textual modalities within a unified autoregressive framework.<n>ASVR improves LLaVA-1.5 by 5% in average scores across 14 multimodal benchmarks.
  arXiv  Detail & Related papers  (2025-06-10T17:57:50Z)
• From Generation to Generalization: Emergent Few-Shot Learning in Video Diffusion Models [65.0487600936788]
  Video Diffusion Models (VDMs) have emerged as powerful generative tools capable of synthesizing high-quality content.<n>We argue that VDMs naturally push to probe structured representations and an implicit understanding of the visual world.<n>Our method transforms each task into a visual transition, enabling the training of LoRA weights on short input- sequences.
  arXiv  Detail & Related papers  (2025-06-08T20:52:34Z)
• Vision-Language Model Dialog Games for Self-Improvement [7.601039502641712]
  VLM Dialog Games is a novel and scalable self-improvement framework for vision-language models (VLMs)<n>By filtering for successful game interactions, we automatically curate a high-quality dataset of interleaved images and text.<n>We demonstrate that fine-tuning on this synthetic data leads to performance gains on downstream tasks and generalises across datasets.
  arXiv  Detail & Related papers  (2025-02-04T21:58:07Z)
• COSMOS: Cross-Modality Self-Distillation for Vision Language Pre-training [49.2684130383925]
  We propose COSMOS: CrOSs-MOdality Self-distillation for vision-language pre-training.<n>It integrates a novel text-cropping strategy and cross-attention module into a self-supervised learning framework.<n>It consistently outperforms previous strong baselines on various zero-shot downstream tasks.
  arXiv  Detail & Related papers  (2024-12-02T18:56:06Z)
• Cambrian-1: A Fully Open, Vision-Centric Exploration of Multimodal LLMs [61.143381152739046]
  We introduce Cambrian-1, a family of multimodal LLMs (MLLMs) designed with a vision-centric approach.<n>Our study uses LLMs and visual instruction tuning as an interface to evaluate various visual representations.<n>We provide model weights, code, supporting tools, datasets, and detailed instruction-tuning and evaluation recipes.
  arXiv  Detail & Related papers  (2024-06-24T17:59:42Z)
• Cross-Modal Self-Training: Aligning Images and Pointclouds to Learn Classification without Labels [69.55622471172941]
  Large-scale vision 2D vision language models, such as CLIP can be aligned with a 3D encoder to learn generalizable (open-vocabulary) 3D vision models. We propose an optimization framework: Cross-MoST: Cross-Modal Self-Training, to improve the label-free classification performance of a zero-shot 3D vision model.
  arXiv  Detail & Related papers  (2024-04-15T21:30:50Z)
• Heuristic Vision Pre-Training with Self-Supervised and Supervised Multi-Task Learning [0.0]
  We propose a novel pre-training framework by adopting both self-supervised and supervised visual pre-text tasks in a multi-task manner. Results show that our pre-trained models can deliver results on par with or better than state-of-the-art (SOTA) results on multiple visual tasks.
  arXiv  Detail & Related papers  (2023-10-11T14:06:04Z)

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.