OSWorld-Science: A Benchmark of Computer Use Agents for Learning and Using Scientific Software

2026-09-30

AI Summary

This study addresses the lack of systematic evaluation regarding the capabilities of Vision-Language Model (VLM) agents in complex scientific software workflows. To this end, we construct a benchmark encompassing multi-domain scientific tasks through expert-informed and iterative co-design methodologies. Furthermore, we propose an artifact-based evaluation mechanism alongside a specialized agent framework that facilitates fine-grained partial scoring and execution trajectory analysis. Our experiments reveal that current state-of-the-art models still encounter substantial challenges when handling scientific tasks. By establishing critical evaluation dimensions, this work provides foundational guidance for the future development of computer-use agents within research contexts.

Abstract

Scientific software presents a demanding test for computer-using agents based on visual language models (VLMs): completing a research workflow requires interpreting specialized interfaces, manipulating scientific objects, and producing verifiable results. We thus introduce OSWorld-Science, a benchmark and evaluation environment that combines scientifically meaningful tasks, artifact-based evaluation, and an efficient agent harness for studying computer use in the scientific domain. The benchmark contains 12 VLMs and 146 high-quality tasks across several scientific domains and software configurations, covering workflows such as molecular drawing and retrosynthesis, pathology image analysis, statistical computing, and physical simulation. Tasks are developed through expert proposals and iterative human--AI co-design, with selection guided by scientific value and difficulty. Task-specific execution-based evaluators inspect application states and generated artifacts, including molecular structures, segmentation masks, plots, and numerical results, and award partial credit for incomplete outcomes. Our special harness integrates model adapters, interaction-loop control, and trajectory logging to support comparisons of models and interaction strategies. Our results show that current state-of-the-art VLMs with a strong harness still face challenges in addressing key questions in the scientific domains. We also analyze the benchmarking results across multi-linguistics, reasoning efforts, context length and other factors and derive several important conclusions and directions to assist future development. Overall, we provide an integrated framework connecting expert-defined scientific goals to verifiable software outcomes, enabling systematic evaluation of both agent capabilities and harness design in scientific workflows.

Problem

• Computer Use Agents
• Visual Language Models
• Scientific Software
• Benchmark
• Research Workflows

Innovation

• Scientific Software Benchmark
• Visual Language Models
• Artifact-based Evaluation
• Computer Use Agents
• Agent Harness