AIM-Bench (Affective Image Manipulation) - Positive Emotion Accuracy: leaderboard
Metric: Emotion accuracy (%) on the items whose target is a positive emotion (amusement, awe, contentment, excitement): Gemini-2.5-Pro classifies each edited image into the 8 Mikels emotions, over the 800 AIM-Bench affective image editing items (EmoSet source images, a target Mikels emotion with target valence-arousal-dominance coordinates and an editing instruction; 8 emotion categories, 5 editing types), each model run once with its official default settings; higher is better. Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 13 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen-Image-Edit-2509 | 73.87 |
| 2 | Step1X-Edit (AIM-Bench checkpoint unspecified) | 67.59 |
| 3 | Seedream 4.0 | 66.32 |
| 4 | Qwen-Image-Edit-Plus (AIM-Bench checkpoint unspecified) | 62.94 |
| 5 | UniWorld-V2 | 62.66 |
| 6 | OmniGen2 | 61.65 |
| 7 | Flux-kontext-pro | 61.01 |
| 8 | Flux-kontext-max | 58.94 |
| 9 | BAGEL-7B-MoT | 55.39 |
| 10 | SeedEdit 3.0 | 54.87 |
Interactive version: theaggregate.ai/benchmark?slug=aim-bench-affective-image-manipulation-positive-emotion-accuracy · How It Works · Data refreshed daily, snapshot 2026-10-07.