Search papers, labs, and topics across Lattice.
OmniVCBench is a new benchmark designed to evaluate the evidence-grounded reasoning capabilities of AI Virtual Cells (AIVCs), focusing on their ability to interpret experimental figures and formulate biological hypotheses. It comprises 6,077 figure-centric question-answer pairs covering prediction, explanation, and discovery tasks, drawing from scientific literature and guided by Bloom's taxonomy. The benchmark also introduces AIVC-Judge, an MLLM-based evaluation framework, and a hard-negative mining strategy for efficient assessment.
A single benchmark now rigorously tests how AI models interpret scientific figures to predict, explain, and discover biological mechanisms, moving beyond mere simulation.
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.
