Text-to-Image Generation on Multi-category Instance Composition Prompt Set
86.44OAELDiff
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| ELDiffBackbone=SD v3.5, Finetuning Strategy=ft. w. Ours2026.06 | 86.44 | 99.97 | 99.64 | 81.71 | |
| TokenComposeBackbone=SD v3.5, Finetuning Strategy=ft. w. TokenCompose2026.06 | 82.33 | 99.97 | 98.39 | 75.69 | |
| IterCompBackbone=SD v3.5, Finetuning Strategy=ft. w. IterComp2026.06 | 82.08 | 99.97 | 98.14 | 76.44 | |
| CoMatBackbone=SD v3.5, Finetuning Strategy=ft. w. CoMat2026.06 | 81.22 | 99.97 | 97.33 | 74.56 | |
| SD v3.5Mode=frozen2026.06 | 80.84 | 99.96 | 96.96 | 73.28 | |
| LLDMBackbone=SD v3.5, Finetuning Strategy=ft. w. LLDM2026.06 | 80.1 | 99.96 | 97.02 | 73.45 | |
| ELDiffBackbone=SD v2.1, Finetuning Strategy=ft. w. Ours2026.06 | 71.16 | 88.7 | 54.01 | 13.92 | |
| IterCompBackbone=SD v2.1, Finetuning Strategy=ft. w. IterComp2026.06 | 63.41 | 80.72 | 35.44 | 5.43 | |
| TokenComposeBackbone=SD v2.1, Finetuning Strategy=ft. w. TokenCompose2026.06 | 60.1 | 80.48 | 36.69 | 5.71 | |
| CoMatBackbone=SD v2.1, Finetuning Strategy=ft. w. CoMat2026.06 | 58.44 | 78.58 | 32.79 | 4.96 | |
| LLDMBackbone=SD v2.1, Finetuning Strategy=ft. w. LLDM2026.06 | 55.09 | 76.43 | 32.07 | 4.73 | |
| SD v2.1Mode=frozen2026.06 | 47.28 | 70.14 | 25.57 | 3.27 |