Skip to main content

Science

Diagram Understanding (VLM diagram comprehension)

Diagram-understanding test suite for large vision-language models: multiple-choice questions about entities and relationships over synthetic icon-graph diagrams and real diagrams (physics, ecology, astronomy, geology, developmental biology, geochemistry). Per-item correctness for GPT-4o, GPT-4V and Gemini-1.5 under base and CoT prompting.

32,764items
3subjects
100%observed
unknownlicense
sciencedomain
generaldomain
imagemodality
textmodality

Response matrix

Fit to width. Hover for subject & item; click a cell for details.

This condition combination wasn’t evaluated — try a different attack, category, or judge.
Correct (1)Incorrect (0)Unobserved

Scale: 1 = correct · 0 = incorrect

Subjects

  1. 1GPT-4o0.731
  2. 2Gemini 1.5 Pro0.6971
  3. 3GPT-4V0.6698