Generative Modeling on TCF
181.7χ²_PQMFPFT
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| FPFTDescription=Tadpole-B fine-tuned with FPFT, Rel. Time=1.00, Trainable Params=12.3M2026.05 | 181.7 | 0.47 | 0.13 | 0.76 | 20.16 | |
| LoRA 32Description=Tadpole-B fine-tuned with LoRA-322026.05 | 256.5 | 0.81 | 0.47 | 2.28 | 32.06 | |
| AFNORel. Time=1.81, Trainable Params=64.1M2026.05 | 373.9 | 3.37 | 18.6 | 5.35 | 274.19 | |
| UNetGenCFDRel. Time=183.72, Trainable Params=100.0M2026.05 | 395.5 | 0.63 | 0.74 | 11.49 | 26.81 | |
| AViTRel. Time=10.39, Trainable Params=60.0M2026.05 | 557.78 | 36.53 | 19.45 | 12.57 | 1,013.51 | |
| ScratchDescription=Tadpole-B trained from scratch2026.05 | 1,775.7 | 7.08 | 25.32 | 40.11 | 37.3 |