VisualEvolve: A Skill Self-Evolution Framework for Visual Reasoning
To address these challenges, we propose VisualEvolve, a self-evolving framework that evolves executable visual operations and tool-use strategies for VLMs through a Select--Execute--Evolve--Evaluate cycle.
ProofPaper ↗
Key points
- Visual tool use enables vision-language models (VLMs) to acquire task-relevant evidence, but its effectiveness can be limited by predefined visual operations and the tool-use strategies learned during training.
- However, skill refinement can introduce performance regressions.
- Skill-Library Snapshot Selection uses performance on a fixed validation set disjoint from execution tasks to select an effective historical snapshot for subsequent evolution, limiting the propagation of performance regressions.
- Moreover, VisualEvolve enables effective skill self-evolution with over an order of magnitude fewer task-solving trajectories than estimated for representative RL-based methods.
Sources (1)
- [1]VisualEvolve: A Skill Self-Evolution Framework for Visual ReasoningarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 20, 03:37 PM
To address these challenges, we propose VisualEvolve, a self-evolving framework that evolves executable visual operations and tool-use strategies for VLMs through a Select--Execute--Evolve--Evaluate cycle.
Visual tool use enables vision-language models (VLMs) to acquire task-relevant evidence, but its effectiveness can be limited by predefined visual operations and the tool-use strategies learned during training.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 20, 2026Voxtral 3
- Oct 20, 2026Do Diffusion LLMs Scale? Testing Mercury 3 + Decide
- Oct 20, 2026Qwen-Audio-3.1-ASR Technical Report
- Oct 19, 2026TokenGate: Adapting LLMs for Fast and Memory-Efficient Prefill with Learned Token Pruning
- Oct 19, 2026OpenVLHarness: Real World Visual Agentic Tool-Use with Multimodal Memory
- Oct 19, 2026Agent World Model in the Wild: Computer-Use Agent Environment Synthesis