VMetaphor-Bench - Dimension Score (Descriptive Prompts): leaderboard
VMetaphor-Bench (Chen et al., 2026) with descriptive prompts that spell out how each visual metaphor should look. A Qwen3.5-27B judge rates each generated image from 1 to 5 on metaphoric efficacy, metaphor logic and perceptual harmony, and the board is the mean of the three ratings; higher is better.
Metric: Judge rating (1-5) (Qwen3.5-27B rates each generated image from 1 to 5 on metaphoric efficacy, metaphor logic and perceptual harmony against the structured annotation, and the board takes the mean of the three; 1,500 visual metaphors; descriptive prompts that spell out the visual realization). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT Image 1.5 | 4.24 | |
| 2 | Nano Banana 2 (Gemini 3.1 Flash Image Preview) | 4.18 | |
| 3 | Seedream 5.0 Lite | 4.15 | |
| 4 | FLUX.2-dev | 4.09 | |
| 5 | Qwen-Image-2512 | 4.03 | |
| 6 | Z-Image-Turbo | 3.9 | |
| 7 | FLUX.1-dev | 3.66 | |
| 8 | Stable Diffusion 3.5 Large | 3.54 | |
| 9 | Janus-Pro-7B | 3.48 | #7341 |
| 10 | BAGEL-7B-MoT | 3.47 | #5515 |
Interactive version: theaggregate.ai/benchmark?slug=vmetaphor-bench-dimension-score-descriptive-prompts · How It Works · Data refreshed daily, snapshot 2026-11-12.