AIM-Bench (Affective Image Manipulation) - Semantic Consistency: leaderboard

AIM-Bench affective image editing scored by the VIEScore semantic consistency rating of how well the edit follows the instruction, on a 0 to 10 scale, higher is better.

Metric: VIEScore semantic consistency (SC, 0-10): how well the edited image follows the editing instruction, 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. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Seedream 4.08.28
2Qwen-Image-Edit-25098.13
3SeedEdit 3.07.85
4Qwen-Image-Edit-Plus (AIM-Bench checkpoint unspecified)7.74
5FLUX-Kontext-Max7.72
6UniWorld-V27.66
7FLUX-Kontext-Pro7.52
8Step1X-Edit (AIM-Bench checkpoint unspecified)7.52
9FLUX.1-Kontext-dev7.14
10OmniGen26.74#4579
11BAGEL-7B-MoT6.47#4619
12DreamOmni26.02
13InstructPix2Pix3.69

Interactive version: theaggregate.ai/benchmark?slug=aim-bench-affective-image-manipulation-semantic-consistency · How It Works · Data refreshed daily, snapshot 2026-11-21.