SciFigQual-Bench: leaderboard
Metric: Mean absolute error of the overall figure-quality score against the mean expert score (1-10 rating scale, lower is better; 1,200 figures from 254 ACL, EMNLP, ICML and NeurIPS papers, each rated with its caption and citing paragraphs; one direct VLM call per figure). Source: arxiv.org. Saturation forecast: Estimated already saturated. 11 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 4.8 | 0.44 |
| 2 | GPT-5.6 Sol | 0.45 |
| 3 | Gemini 3.5 Flash | 0.46 |
| 4 | Nova Pro | 0.47 |
| 5 | Claude Sonnet 5 | 0.48 |
| 6 | Llama 4 Maverick | 0.49 |
| 7 | Pixtral Large | 0.5 |
| 8 | InternVL3-78B | 0.54 |
| 9 | Seed 2.0 Pro | 0.61 |
| 10 | GLM-4.6V | 0.66 |
Interactive version: theaggregate.ai/benchmark?slug=scifigqual-bench · How It Works · Data refreshed daily, snapshot 2026-09-29.