VMetaphor-Bench - Structure Type: leaderboard

Metric: MCQ accuracy (%; one question per image on how the two domains are combined: replacement, fusion or juxtaposition; the generated image is judged by Qwen3.5-27B on multiple-choice questions (four options including cannot be determined) built from each metaphor's structured annotation; 1,500 visual metaphors curated from real creative imagery; conceptual prompts that state only the metaphorical idea). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 11 models tracked.

Top models

#ModelScore
1GPT Image 1.578.1
2Nano Banana 2 (Gemini 3.1 Flash Image Preview)76.3
3Qwen-Image-251266.1
4Seedream 5.0 Lite65.6
5FLUX.2-dev63.7
6Stable Diffusion 3.5 Large60.1
7Z-Image-Turbo59.6
8Janus-Pro-7B59.1
9FLUX.1-dev54.9
10BAGEL-7B-MoT54.8

Interactive version: theaggregate.ai/benchmark?slug=vmetaphor-bench-structure-type · How It Works · Data refreshed daily, snapshot 2026-09-26.