Scientific problem-solving involves synthesizing information while applying expert knowledge. We introduce CURIE, a scientific long-Context Understanding,Reasoning and Information Extraction benchmark to measure the potential of Large Language Models (LLMs) in scientific problem-solving and assisting scientists in realistic workflows. This benchmark introduces ten challenging tasks with a total of 580 problems and solution pairs curated by experts in six disciplines - materials science, condensed matter physics, quantum computing, geospatial analysis, biodiversity, and proteins - covering both experimental and theoretical work-flows in science. We evaluate a range of closed and open LLMs on tasks in CURIE which requires domain expertise, comprehension of long in-context information,and multi-step reasoning. While Gemini Flash 2.0 and Claude-3 show consistent high comprehension across domains, the popular GPT-4o and command-R+ fail dramatically on protein sequencing tasks. With the best performance at 32% there is much room for improvement for all models. We hope that insights gained from CURIE can guide the future development of LLMs in sciences. Evaluation code and data are inthis https URL
View on arXiv@article{cui2025_2503.13517,
title={ CURIE: Evaluating LLMs On Multitask Scientific Long Context Understanding and Reasoning },
author={ Hao Cui and Zahra Shamsi and Gowoon Cheon and Xuejian Ma and Shutong Li and Maria Tikhanovskaya and Peter Norgaard and Nayantara Mudur and Martyna Plomecka and Paul Raccuglia and Yasaman Bahri and Victor V. Albert and Pranesh Srinivasan and Haining Pan and Philippe Faist and Brian Rohr and Ekin Dogus Cubuk and Muratahan Aykol and Amil Merchant and Michael J. Statt and Dan Morris and Drew Purves and Elise Kleeman and Ruth Alcantara and Matthew Abraham and Muqthar Mohammad and Ean Phing VanLee and Chenfei Jiang and Elizabeth Dorfman and Eun-Ah Kim and Michael P Brenner and Viren Jain and Sameera Ponda and Subhashini Venugopalan },
journal={arXiv preprint arXiv:2503.13517},
year={ 2025 }
}