Text-to-Image Generation on sDCI (val)
24.1FIDRT-Lynx
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| RT-LynxModel=Qwen-Image-2512, Strategy=SOTA sparse-weight2026.05 | 24.1 | 1.132 | 26.07 | 92.85 | |
| FullModel=Qwen-Image-2512, Sparsity=None2026.05 | 24.42 | 1.13 | 25.91 | 94.35 | |
| SlimModel=Qwen-Image-2512, Strategy=SOTA sparse-weight2026.05 | 30.22 | 1.103 | 25.83 | 92.79 | |
| Sparse ActivationModel=Qwen-Image-2512, Strategy=naive 2:4 sparsity2026.05 | 51.41 | 0.3205 | 26.48 | 65.93 | |
| BaWAModel=Qwen-Image-2512, Strategy=SOTA sparse-weight2026.05 | 56.33 | -0.2107 | 25.68 | 57.57 | |
| WandaModel=Qwen-Image-2512, Strategy=SOTA sparse-weight2026.05 | 59.55 | -0.2371 | 25.54 | 54.61 | |
| RIAModel=Qwen-Image-2512, Strategy=SOTA sparse-weight2026.05 | 62.81 | -0.3143 | 25.59 | 54.05 | |
| Sparse WeightModel=Qwen-Image-2512, Strategy=naive 2:4 sparsity2026.05 | 119.8 | -1.512 | 20.276 | 28.4 |