OmniVCBench: Benchmarking Evidence-Grounded Multimodal Reasoning Towards AI Virtual Cells
cs.AI, q-bio.QM
Submitted: 2026-09-29
Updated: 2026-09-29
Comments: 45 pages, 16 figures;
Code: https://github.com/MiniMax-AI/MiniMax-M3
License: http://creativecommons.org/licenses/by/4.0/
The gist: Artificial Intelligence Virtual Cells (AIVCs) are envisioned as scientific agents that simulate cellular responses, explain underlying mechanisms, and support hypothesis-driven discovery.
Terminology
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.
Sources
- Qwen3-VL Technical Report
- Qwen2.5-VL Technical Report
- HuatuoGPT-Vision, Towards Injecting Medical Visual Knowledge into Multimodal LLMs at Scale
- GLM-5V-Turbo: Toward a Native Foundation Model for Multimodal Agents
- LoRA: Low-Rank Adaptation of Large Language Models
- LABBench2: An Improved Benchmark for AI Systems Performing Biology Research
- LLaVA-OneVision: Easy Visual Task Transfer
- Qwen3-VL-Embedding and Qwen3-VL-Reranker: A Unified Framework for State-of-the-Art Multimodal Retrieval and Ranking
- MMSci: A Dataset for Graduate-Level Multi-Discipline Multimodal Scientific Understanding
- Benchmarking virtual cell models for in-the-wild perturbation response
- SmolVLM: Redefining small and efficient multimodal models
- Virtual Cells: Predict, Explain, Discover
- InternVL3.5: Advancing Open-Source Multimodal Models in Versatility, Reasoning, and Efficiency
- DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence
- HiSciBench: A Hierarchical Multi-disciplinary Benchmark for Scientific Intelligence from Reading to Discovery
- SC-Arena: A Natural Language Benchmark for Single-Cell Reasoning with Knowledge-Augmented Evaluation
Related papers
- MAVEN-T: Reinforced Heterogeneous Distillation for Real-Time Multi-Agent Trajectory Prediction
- Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
- Agents' Last Exam
- The Clinician's Veto: Navigating Trust, Liability, and Uncertainty in Autonomous AI Prescribing
- MindHelper: Closed-Loop Embodied Mental-State Reasoning for Precision Intervention
- Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems