S1-VL: Scientific Multimodal Reasoning Model with Thinking-with-Images

TL;DR

S1-VL is a multimodal reasoning model for scientific domains that combines structured chain-of-thought reasoning with active image manipulation via code execution, achieving state-of-the-art results on multiple benchmarks.

Contribution

The paper introduces S1-VL, a novel scientific multimodal reasoning model supporting both reasoning paradigms and a new data filtering strategy for effective training.

Findings

S1-VL-32B achieves state-of-the-art performance on all five Thinking-with-Images benchmarks.

The model outperforms existing systems on scientific reasoning benchmarks such as Physics and VRSBench.

The multi-stage filtering pipeline improves the quality of training data by reducing ineffective visual operations.

Abstract

We present S1-VL, a multimodal reasoning model for scientific domains that natively supports two complementary reasoning paradigms: Scientific Reasoning, which relies on structured chain-of-thought, and Thinking-with-Images, which enables the model to actively manipulate images through Python code execution during reasoning. In the Thinking-with-Images mode, the model generates and executes image-processing code in a sandbox environment, obtains intermediate visual results, and continues reasoning in a multi-turn iterative manner. This design is particularly effective for challenging scenarios such as high-resolution scientific chart interpretation, microscopic image understanding, and geometry-assisted reasoning. To construct the training data, we collect scientific multimodal datasets spanning six disciplines: mathematics, physics, chemistry, astronomy, geography, and biology. We further develop a six-dimensional quality filtering framework for reasoning trajectories. To mitigate redundant, ineffective, and erroneous visual operations commonly found in existing datasets, we propose a multi-stage filtering pipeline together with an adaptive data routing strategy. This strategy converts samples with low visual information gain into pure Reasoning-mode data, enabling the model to learn when image operations are truly necessary. S1-VL is trained through a four-stage progressive pipeline: scientific multimodal SFT, Thinking-with-Images cold-start SFT, and two stages of reinforcement learning with SAPO. We build S1-VL-32B on top of Qwen3-VL-32B-Thinking and evaluate it on 13 benchmarks. Experimental results show that S1-VL-32B achieves state-of-the-art performance on all five Thinking-with-Images benchmarks, including HRBench-4K, HRBench-8K, MME-RealWorld-CN, MME-RealWorld-Lite, and V*, and outperforms compared systems on scientific reasoning benchmarks such as Physics and VRSBench.