Compound AI System Alignment on Coherent and Progressive Image Sequence Generation (test)
0.25Preference ScoreSysDPO-Direct
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SysDPO-DirectMode=Proposed2025.02 | 0.25 | 73 | |
| Only Train Language ModelLanguage Model=Trained, Diffusion Model=Frozen2025.02 | 0.23 | 65 | |
| Best-of-SamplingSample Size=42025.02 | 0.16 | 67 | |
| Only Train Diffusion ModelLanguage Model=Frozen, Diffusion Model=Trained2025.02 | -0.03 | 38 | |
| System Before AlignmentBackbone=Llama-3-8B2025.02 | -0.2 | 32 |