ChemArgus benchmark diagnoses detailed reasoning steps in chemistry problems
ChemArgus: Automated Scientific Diagnostics for Lab-Grade Multimodal Chemistry Reasoning: A Frontier Benchmark
Machine Learning
Summary
Chemistry tests usually only check if an answer is right or wrong, missing how people or AI come to that answer. The authors created ChemArgus, a new test that scores every step in a chemistry problem to see exactly where a model gets stuck. It also compares how well the model understands pictures and text in the same question. This helps show not just if an AI can answer but where its thinking breaks down.
What this means in practice
- •For ai developers for scientific applications: Measure and diagnose where AI models fail in solving complex chemistry problems step-by-step, improving targeted training approaches.
- •For educational software developers: Create teaching tools that highlight mistakes in chemistry problem solving by scoring partial steps and comparing visual and textual understanding.
Authors
Hanyao Guo, Mo Cui, Zihan Tan, Leixin Sun, Yichun Wang, Yinan Shen, Guojun Zhu, Jia Jiang, Erhao Chen, Zui Chen, Wenzhe Gao, Zhihao Yu, Qijun Wang, Yetian Zhu, Jiande Chen, Chuhang Pei, Yao Jiang, Yichen Fan, Pinhan Wang, Xiaorui Pan, Qianyi Zhu, Yang Liu, Mang Ye, Guancheng Wan
Abstract
Chemistry evaluation has moved past multiple-choice probes toward open-ended expert problems, yet the benchmarks that define the current frontier still score answer endpoints alone. Such scoring conflates memorized facts with stepwise derivation, treats visual and textual inputs as interchangeable, and confounds long-horizon planning with stepwise execution. These blind spots prevent any diagnosis of where the reasoning chain breaks in a model. We present ChemArgus, a competition-level chemistry benchmark drawn from chemistry competition material, textbooks, and research papers, whose scoring replaces the answer endpoint with three measurement properties. Fine-grained scoring grades every subquestion point by point with partial credit, so a score lands at the level of the step rather than the outcome. Multimodal capability is priced on every problem containing a real image, paired with a controlled text description, so visual understanding is measured against text on identical tasks. Diagnostic attribution reads paired metrics that share one scoring unit and differ only in context construction, so a gap between two scores isolates a single cause, namely integration, propagation, or local reasoning. GPT-5.6-Sol scores 23.2% on unassisted whole-problem solving against 28.8% and 30.4% under the two sequential metrics, a gap measuring the integration burden. Through rubric partial credit and these cause-isolating score gaps, ChemArgus shows not only whether a model answers chemistry correctly but where its reasoning chain fails, so that every score it reports doubles as a diagnosis, a readout of the chain rather than a verdict, aiming to guide language models step by step on their path toward chemistry expertise, one corrected step at a time, each step priced by the rubric, point by point.