From Macro to Micro: Benchmarking Microscopic Spatial Intelligence on Molecules via Vision-Language Models

TL;DR

This paper introduces MiSI-Bench, a large-scale benchmark for evaluating Vision-Language Models on microscopic spatial reasoning tasks involving molecules, revealing current models' limitations and potential for domain-specific improvement.

Contribution

The paper presents a new benchmark dataset and evaluation framework for microscopic spatial intelligence, enabling systematic assessment of VLMs in scientific molecular reasoning tasks.

Findings

Current VLMs perform below human level on the benchmark.

A fine-tuned 7B model surpasses humans in spatial transformation tasks.

Explicit domain knowledge is crucial for scientific reasoning in VLMs.

Abstract

This paper introduces the concept of Microscopic Spatial Intelligence (MiSI), the capability to perceive and reason about the spatial relationships of invisible microscopic entities, which is fundamental to scientific discovery. To assess the potential of Vision-Language Models (VLMs) in this domain, we propose a systematic benchmark framework MiSI-Bench. This framework features over 163,000 question-answer pairs and 587,000 images derived from approximately 4,000 molecular structures, covering nine complementary tasks that evaluate abilities ranging from elementary spatial transformations to complex relational identifications. Experimental results reveal that current state-of-the-art VLMs perform significantly below human level on this benchmark. However, a fine-tuned 7B model demonstrates substantial potential, even surpassing humans in spatial transformation tasks, while its poor performance in scientifically-grounded tasks like hydrogen bond recognition underscores the necessity of integrating explicit domain knowledge for progress toward scientific AGI. The datasets are available at https://huggingface.co/datasets/zongzhao/MiSI-bench.