SciFig-Bench - Rubric-Based Content Quality: leaderboard
SciFig-Bench (Huang et al., 2026) asks a system to draw the methodology figure of an AI or ML paper from its method text and a figure description, over 435 human-verified papers from 15 venues. Rubric-based content quality (0-10): the mean of technical accuracy, clarity, structural consistency and readability scores from 20 binary rubric questions answered by Claude Opus 4.6, with rubrics derived from held-out real figures; higher is better.
Metric: Rubric-based content quality (0-10) (mean of technical accuracy, clarity, structural consistency and readability, 20 binary rubric questions answered by Claude Opus 4.6). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 3 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Nano Banana Pro | 7.5 | |
| 2 | SciFig GPT-5.4-Image (checkpoint unspecified) | 6.1 | |
| 3 | Qwen-Image | 3.9 |
Interactive version: theaggregate.ai/benchmark?slug=scifig-bench-rubric-based-content-quality · How It Works · Data refreshed daily, snapshot 2026-11-12.