Decoding Scientific Experimental Images: The SPUR Benchmark for Perception, Understanding, and Reasoning

arXiv cs.CV / 5/1/2026

📰 NewsIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper introduces SPUR, a new benchmark for scientific experimental image perception, understanding, and reasoning, built from 1,084 expert-curated images and 4,264 QA pairs.
SPUR evaluates multimodal LLMs using panel-level fine-grained perception across numerical, morphological, and information-localization dimensions on six panel types.
The benchmark measures cross-panel relation understanding by using complex samples with an average of 14.3 panels per image.
It also tests expert-level qualitative and quantitative reasoning across five experimental paradigms to see whether models can draw conclusions from evidence.
Experiments on 20 MLLMs and four multimodal Chain-of-Thought methods show substantial gaps versus expert performance, highlighting a key bottleneck for AI for Science (AI4S).

Abstract

We introduce SPUR, a comprehensive benchmark for scientific experimental image perception, understanding, and reasoning, comprising 4,264 question-answering (QA) pairs derived from 1,084 expert-curated images. SPUR features three key innovations: (1) Panel-Level Fine-Grained Perception: evaluating the visual perception of multimodal large language models (MLLMs) across three dimensions (numerical, morphological, and information localization) on six fine-grained panel types; (2) Cross-Panel Relation Understanding: utilizing complex images with an average of 14.3 panels per sample to evaluate MLLMs' ability to decipher intricate cross-panel relations; (3) Expert-Level Reasoning: assessment of qualitative and quantitative reasoning across five experimental paradigms to determine if models can infer conclusions from evidence as human experts do. Comprehensive evaluation of 20 MLLMs and four multimodal Chain-of-Thought (MCoT) methods reveals that current models fall significantly short of the expert-level requirements for scientific image interpretation, underscoring a critical bottleneck in AI for Science (AI4S) research.