Multi-subject image consistency on MSBench multi-subject
0.749CLIP-INano-Banana
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Nano-BananaCategory=In-context, Inference steps=302026.02 | 0.749 | 0.114 | 0.389 | |
| FluxKontextCategory=In-context, Inference steps=302026.02 | 0.732 | 0.107 | 0.372 | |
| SSR-EncoderCategory=Adapters & MLLM, Inference steps=302026.02 | 0.723 | 0.104 | 0.311 | |
| λ-ECLIPSECategory=Adapters & MLLM, Inference steps=302026.02 | 0.719 | 0.097 | 0.304 | |
| StoryTailorCategory=Proposed, Inference steps=302026.02 | 0.717 | 0.112 | 0.414 | |
| Qwen-EditCategory=In-context, Inference steps=302026.02 | 0.714 | 0.108 | 0.396 | |
| MS-DiffusionCategory=Adapters & MLLM, Inference steps=302026.02 | 0.692 | 0.108 | 0.34 |