# Vero: An Open RL Recipe for General Visual Reasoning

arXiv: https://arxiv.org/abs/2604.04917 · HTML: https://mlsanity.com/paper/2604.04917/

## Key facts

- Cited by 11 papers in the map (data through 14 Oct 2026).
- Placed in Multimodal reasoning, one of 17 directions in LLM reasoning.
- Builds on 33 of its 115 references that are in the map.
- First posted on arXiv 6 Apr 2026.

## Placed in

- LLM reasoning / [Multimodal reasoning](https://mlsanity.com/lr/multimodal-reasoning/)

## Builds on (most cited first, 30 of 33)

- [Proximal Policy Optimization Algorithms](https://mlsanity.com/paper/1707.06347/) (2017)
- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [Qwen2.5-VL Technical Report](https://mlsanity.com/paper/2502.13923/) (2025)
- [Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities](https://mlsanity.com/paper/2507.06261/) (2025)
- [Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context](https://mlsanity.com/paper/2403.05530/) (2024)
- [Qwen3-VL Technical Report](https://mlsanity.com/paper/2511.21631/) (2025)
- [OpenAI GPT-5 System Card](https://mlsanity.com/paper/2601.03267/) (2025)
- [Gemini: A Family of Highly Capable Multimodal Models](https://mlsanity.com/paper/2312.11805/) (2023)
- [Group Sequence Policy Optimization](https://mlsanity.com/paper/2507.18071/) (2025)
- [Kimi K2.5: Visual Agentic Intelligence](https://mlsanity.com/paper/2602.02276/) (2026)
- [PathVQA: 30000+ Questions for Medical Visual Question Answering](https://mlsanity.com/paper/2003.10286/) (2020)
- [GLM-4.5V and GLM-4.1V-Thinking: Towards Versatile Multimodal Reasoning with Scalable Reinforcement Learning](https://mlsanity.com/paper/2507.01006/) (2025)
- [Kimi-VL Technical Report](https://mlsanity.com/paper/2504.07491/) (2025)
- [MM-Eureka: Exploring the Frontiers of Multimodal Reasoning with Rule-based Reinforcement Learning](https://mlsanity.com/paper/2503.07365/) (2025)
- [LLaVA-OneVision-1.5: Fully Open Framework for Democratized Multimodal Training](https://mlsanity.com/paper/2509.23661/) (2025)
- [Molmo2: Open Weights and Data for Vision-Language Models with Video Understanding and Grounding](https://mlsanity.com/paper/2601.10611/) (2026)
- [Olmo 3](https://mlsanity.com/paper/2512.13961/) (2025)
- [MiMo-VL Technical Report](https://mlsanity.com/paper/2506.03569/) (2025)
- [GeomVerse: A Systematic Evaluation of Large Models for Geometric Reasoning](https://mlsanity.com/paper/2312.12241/) (2023)
- [Gemini Robotics: Bringing AI into the Physical World](https://mlsanity.com/paper/2503.20020/) (2025)
- [V-STaR: Benchmarking Video-LLMs on Video Spatio-Temporal Reasoning](https://mlsanity.com/paper/2503.11495/) (2025)
- [Defeating the Training-Inference Mismatch via FP16](https://mlsanity.com/paper/2510.26788/) (2025)
- [MM-Vet v2: A Challenging Benchmark to Evaluate Large Multimodal Models for Integrated Capabilities](https://mlsanity.com/paper/2408.00765/) (2024)
- [OneThinker: All-in-one Reasoning Model for Image and Video](https://mlsanity.com/paper/2512.03043/) (2025)
- [FineVision: Open Data Is All You Need](https://mlsanity.com/paper/2510.17269/) (2025)
- [Ray Interference: a Source of Plateaus in Deep Reinforcement Learning](https://mlsanity.com/paper/1904.11455/) (2019)
- [Game-RL: Synthesizing Multimodal Verifiable Game Data to Boost VLMs' General Reasoning](https://mlsanity.com/paper/2505.13886/) (2025)
- [We-Math 2.0: A Versatile MathBook System for Incentivizing Visual Mathematical Reasoning](https://mlsanity.com/paper/2508.10433/) (2025)
- [Metacognitive Reuse: Turning Recurring LLM Reasoning Into Concise Behaviors](https://mlsanity.com/paper/2509.13237/) (2025)
- [Visual Jigsaw Post-Training Improves MLLMs](https://mlsanity.com/paper/2509.25190/) (2025)

## Built on by (most cited first, 11 of 11)

- [When Teachers Mislead: Spurious-Signal-Aware On-Policy Distillation](https://mlsanity.com/paper/2608.03632/) (2026)
- [Veritas++: Value-aware On-Policy Distillation for Perception-Enhanced AIGI Detection](https://mlsanity.com/paper/2607.27113/) (2026)
- [Trace: A Taxonomy-Guided Environment for Multidomain Visual Reasoning](https://mlsanity.com/paper/2607.19790/) (2026)
- [MIRROR: Learning from the Other View for Multi-Modal Reasoning](https://mlsanity.com/paper/2607.21552/) (2026)
- [Beacon: Knowing When and How to Perform Agentic Visual Reasoning](https://mlsanity.com/paper/2607.28595/) (2026)
- [OpenVisTool: An Open Recipe for Synthesizing Instructive Visual Tool-Use Trajectories](https://mlsanity.com/paper/2608.08557/) (2026)
- [MMVistaReason: Toward Open-Data and Post-Training Recipes for Multimodal Reasoning](https://mlsanity.com/paper/2610.01352/) (2026)
- [VisionPsy-Nano: Improving Accuracy, Efficiency, and Reliability in On-Device Vision-Language Models](https://mlsanity.com/paper/2609.31746/) (2026)
- [EviViT: Evidence-Adaptive Vision Transformers for Fine-Grained Perception](https://mlsanity.com/paper/2609.37123/) (2026)
- [Harnessing Domain Specialists in Multimodal Mixture-of-Experts for Efficient Adaptation](https://mlsanity.com/paper/2610.02123/) (2026)
- [Trinity: Self-Evolving Vision-Language Models with a Self-Verifier](https://mlsanity.com/paper/2610.04469/) (2026)

## Abstract (arXiv)

What does it take to build a visual reasoner that works across charts, science, spatial understanding, and open-ended tasks? The strongest vision-language models (VLMs) suggest that broad visual reasoning is within reach, yet their closed data and reinforcement learning (RL) pipelines make their gains difficult to study, reproduce, or extend. We introduce Vero, a family of fully open VLMs that match or exceed existing open-weight models across diverse visual reasoning tasks. We scale RL data and rewards across six broad task categories, constructing Vero-600K, a 600K-sample dataset from 59 datasets, and designing task-routed rewards that handle heterogeneous answers. Across VeroEval, our 30-benchmark suite, Vero-600K outperforms existing RL datasets under controlled comparisons. Applied to five starting models, Vero variants gain 2.9-5.4 points on average over their initial models. Notably, Vero-Qwen3I-8B, trained on the Instruct model, surpasses Qwen3-VL-8B-Thinking by 3.8 points on average without additional distillation. Systematic ablations reveal that different task categories elicit distinct reasoning patterns and that broad gains depend on learning them jointly rather than in isolation. All data, code, and models are publicly available.