---
pipeline_tag: image-text-to-text
library_name: transformers
---

Repository for the model used in the paper [SoTA with Less: MCTS-Guided Sample Selection for Data-Efficient Visual Reasoning Self-Improvement](https://huggingface.co/papers/2504.07934).

<div align="center">

<h1>SoTA with Less: MCTS-Guided Sample Selection for Data-Efficient Visual Reasoning Self-Improvement</h1>

<p align="center">
  <a href="https://huggingface.co/russwang/ThinkLite-VL-7B">🤗7B Model</a> • <a href="https://huggingface.co/datasets/russwang/ThinkLite-VL-70k">📊ThinkLite-VL-70k</a> • <a href="https://huggingface.co/datasets/russwang/ThinkLite-VL-hard-11k">📊ThinkLite-VL-Hard-11k</a> • <a href="https://arxiv.org/abs/2504.07934">📄Paper</a>
</p>

</div>

Official codebase of "SoTA with Less: MCTS-Guided Sample Selection for Data-Efficient Visual Reasoning Self-Improvement". We will release all codes in one week.