Omnivcbench benchmarks ai reasoning for interpreting biological figures
OmniVCBench: Benchmarking Evidence-Grounded Multimodal Reasoning Towards AI Virtual Cells
Artificial Intelligence
Summary
Understanding how cells respond to experiments helps scientists discover new biology. The authors created OmniVCBench, a test that checks if AI can accurately interpret scientific images and data to explain and predict cellular behavior. They also introduced AIVC-Judge, a system that evaluates AI answers, and a method to improve multiple-choice questions using tricky errors seen in AI responses. Their work aims to help build AI that supports biological research by reasoning about complex evidence.
What this means in practice
- •For biotech research teams: Assess AI tools that interpret experimental images to help generate accurate biological hypotheses during research.
- •For data science teams in pharma: Evaluate and benchmark AI models that explain cell behavior from scientific literature figures to support drug discovery workflows.
Authors
Manyu Li, Xunkai Li, Yongfu Xiong, Yi Liu, Rong-Hua Li, Guoren Wang
Abstract
Artificial Intelligence Virtual Cells (AIVCs) are envisioned as scientific agents that simulate cellular responses, explain underlying mechanisms, and support hypothesis-driven discovery. Existing AIVC benchmarks, however, operate primarily at the simulation layer, motivating complementary evaluation of how models interpret experimental evidence and formulate biological hypotheses. We introduce OmniVCBench, a figure-centric, source-traceable benchmark for the interpretation component of an AIVC. It contains 6,077 curated single- and multi-subfigure question--answer pairs derived from figures and experimental contexts in the scientific literature. Guided by Bloom's taxonomy, we instantiate interpretation-layer counterparts of the AIVC Predict--Explain--Discover agenda through three scientific reasoning tasks. We further introduce AIVC-Judge, a task-conditioned MLLM-as-a-judge framework with category-specific, reference-aware rubrics for evaluating open-ended responses. A complementary Model-Derived Hard-Negative Mining (MDHNM) strategy converts plausible errors observed during model inference into MCQ distractors for lower-cost evaluation. Within the evaluated heterogeneous model pool, MCQ accuracy correlates positively with AIVC-Judge scores, providing a complementary view of performance alongside open-response evaluation. Code and data demo are available at https://anonymous.4open.science/r/OmniVCBench.