Imaging-101: Benchmarking LLM Coding Agents on Scientific Computational Imaging

DOI10.48550/arXiv.2607.10789MaRDI QIDQ7284032arXivFDO

Authors Yixuan Jiang, Yuxuan Guo, Zhijun Zeng, Tianhan Zhang, He Sun, Siyi Chen, Jiahe Ying, Yixuan Jia, Yuxuan Gu, Enze Ye, Weimin Bai, Shaochi Ren, Binhong Gao, Yubing Li

Publication date 12 July 2026

Published in arXiv

Tldr We present Imaging-101, a benchmark of 57 expert-verified computational imaging tasks across six scientific domains in a standardized four-stage pipeline format for evaluating large language models (LLMs) on their coding capabilities in the field. Evaluating seven frontier LLMs reveals significant challenges that go beyond general coding benchmarks, highlighting a need for domain-specialized agents to aid scientists.

Full work available at URL https://doi.org/10.48550/arXiv.2607.10789

arXiv classification

cs.AI

This page was built for publication: Imaging-101: Benchmarking LLM Coding Agents on Scientific Computational Imaging

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7284032)