Evaluation-driven Scaling for Scientific Discovery
Haotian Ye, Haowei Lin, Jingyi Tang, Yizhen Luo, Caiyin Yang, Chang Su, Rahul Thapa, Rui Yang, Ruihua Liu, Zeyu Li, Chong Gao, Dachao Ding, Guangrong He, Miaolei Zhang, Lina Sun, Wenyang Wang, Yuchen Zhong, Zhuohao Shen, Di He, Jianzhu Ma, Stefano Ermon, Tongyang Li, Xiaowen Chu, James Zou, Yuzhi Xu
Why It Matters
What makes this one worth your time
This work provides a practical framework for researchers and engineers to improve the efficiency and effectiveness of scientific discovery processes using language models.
SimpleTES scales evaluation-driven discovery loops to achieve state-of-the-art solutions in scientific problems.
Summary
The paper introduces Simple Test-time Evaluation-driven Scaling (SimpleTES), a framework that enhances scientific discovery through evaluation-driven loops, demonstrating significant performance improvements across various scientific problems using language models.
Key contributions
- Introduction of the SimpleTES framework for scaling evaluation-driven discovery loops.
- Demonstrated substantial performance improvements across 21 scientific problems.
- Provided trajectory-level histories that enhance feedback-driven learning.
Notable insights
- The integration of feedback-driven refinement with parallel exploration is a strategic approach that may not be widely adopted yet.
- The ability to generalize to unseen problems after post-training on successful trajectories suggests a novel method for enhancing model robustness.
Possible limitations
- Not stated in the abstract.
Abstract
arXiv:2604.19341v1 Announce Type: cross Abstract: Language models are increasingly used in scientific discovery to generate hypotheses, propose candidate solutions, implement systems, and iteratively refine them. At the core of these trial-and-error loops lies evaluation: the process of obtaining feedback on candidate solutions via verifiers, simulators, or task-specific scoring functions. While prior work has highlighted the importance of evaluation, it has not explicitly formulated the problem of how evaluation-driven discovery loops can be scaled up in a principled and effective manner to push the boundaries of scientific discovery, a problem this paper seeks to address. We introduce Simple Test-time Evaluation-driven Scaling (SimpleTES), a general framework that strategically combines parallel exploration, feedback-driven refinement, and local selection, revealing substantial gains unlocked by scaling evaluation-driven discovery loops along the right dimensions. Across 21 scientific problems spanning six domains, SimpleTES discovers state-of-the-art solutions using gpt-oss models, consistently outperforming both frontier-model baselines and sophisticated optimization pipelines. Particularly, we sped up the widely used LASSO algorithm by over 2x, designed quantum circuit routing policies that reduce gate overhead by 24.5%, and discovered new Erdos minimum overlap constructions that surpass the best-known results. Beyond novel discoveries, SimpleTES produces trajectory-level histories that naturally supervise feedback-driven learning. When post-trained on successful trajectories, models not only improve efficiency on seen problems but also generalize to unseen problems, discovering solutions that base models fail to uncover. Together, our results establish effective evaluation-driven loop scaling as a central axis for advancing LLM-driven scientific discovery, and provide a simple yet practical framework for realizing these gains.