TangPoetryBench - Recognizability: leaderboard
TangPoetryBench (Hu et al., 2026) collects human ratings of text-to-image illustrations for 320 classical Chinese Tang poems on ten dimensions, from technical quality and text integrity to core imagery and emotional resonance. This board gives the percentage of annotators who identify the illustrated poem from the image; higher is better.
Metric: Recognizability (%; per-image share of annotators who identify the illustrated poem, averaged over 320 illustrations). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 4 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Seedream 4.5 | 87 | |
| 2 | Nano Banana Pro | 80 | |
| 3 | GPT Image 1 | 79 | |
| 4 | Midjourney V7 | 70.7 |
Interactive version: theaggregate.ai/benchmark?slug=tangpoetrybench-recognizability · How It Works · Data refreshed daily, snapshot 2026-11-12.