Semiosis Quality Evaluation on HGI SemiosisArt
0.79KRCCHuman (Non-expert)
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Human (Non-expert)Group=Human Reference, Backbone=-2026.04 | 0.79 | 0.924 | 0.946 | |
| SemJudge (Gemini-Flash)Group=Semiosis Quality, Backbone=Zero-shot MLLM2026.04 | 0.746 | 0.964 | 0.968 | |
| SemJudge (Qwen-35B-A3B)Group=Semiosis Quality, Backbone=Zero-shot MLLM2026.04 | 0.674 | 0.88 | 0.878 | |
| SemJudge (Qwen-9B)Group=Semiosis Quality, Backbone=Zero-shot MLLM2026.04 | 0.533 | 0.856 | 0.808 | |
| ArtCoTGroup=Formal Analysis, Backbone=Zero-shot MLLM2026.04 | 0.294 | 0.604 | 0.609 | |
| LMM4LMMGroup=Rationale&Scoring, Backbone=Tuned MLLM2026.04 | 0.274 | 0.651 | 0.547 | |
| VIEScoreGroup=Structured Rationale, Backbone=Zero-shot MLLM2026.04 | 0.241 | 0.641 | 0.321 | |
| PickScoreGroup=Preference Scoring, Backbone=CLIP2026.04 | 0.202 | 0.605 | 0.31 | |
| DSGGroup=Structured Rationale, Backbone=Zero-shot MLLM2026.04 | 0.153 | 0.678 | 0.23 | |
| CLIP-IQAGroup=Quality Scoring, Backbone=CLIP2026.04 | 0.08 | 0.25 | 0.088 | |
| ArtiMuseGroup=Art Interpretation, Backbone=Tuned MLLM2026.04 | 0.075 | 0.088 | 0.156 | |
| CLIPScoreGroup=Alignment Scoring, Backbone=CLIP2026.04 | 0.041 | 0.059 | 0.106 | |
| Aesthetic PredictorGroup=Preference Scoring, Backbone=CLIP2026.04 | 0.03 | 0.117 | 0.084 | |
| HPSv2Group=Preference Scoring, Backbone=CLIP2026.04 | 0.03 | -0.017 | -0.081 | |
| DeQA-ScoreGroup=Quality Scoring, Backbone=Tuned MLLM2026.04 | 0.023 | -0.112 | -0.118 | |
| ImageRewardGroup=Preference Scoring, Backbone=BLIP2026.04 | -0.002 | 0.232 | 0.128 | |
| Random GuessGroup=Conventional Scorers, Backbone=-2026.04 | -0.023 | -0.077 | 0.007 | |
| GalleryGPTGroup=Art Interpretation, Backbone=Tuned MLLM2026.04 | -0.034 | -0.201 | -0.087 |