Inconsistency and intrinsic difficulty of different fine-tuning versions of the same generative AI, namely FLUX.l-schnell. Classification error rate and obtained with Dual Vision Transformer —DaViT— model.
| Training Testing | FLUX.1-schnell | Finetuning #1 | Finetuning #2 | Finetuning #3 | Finetuning #4 |
|---|---|---|---|---|---|
| FLUX.l-schnell | 0.91% | 4.44% | 3.42% | 2.23% | 2.52% |
| Finetuning #1 | 2.79% | 1.35% | 2.21% | 2.97% | 3.37% |
| Finetuning #2 | 2.43% | 2.18% | 0.93% | 2.90% | 3.18% |
| Finetuning #3 | 2.61% | 2.80% | 3.75% | 0.98% | 1.81% |
| Finetuning #4 | 3.12% | 3.51% | 2.72% | 1.63% | 0.78% |
| Training Testing | FLUX.1-schnell | Finetuning #1 | Finetuning #2 | Finetuning #3 | Finetuning #4 |
|---|---|---|---|---|---|
| FLUX.l-schnell | 0.91% | 4.44% | 3.42% | 2.23% | 2.52% |
| Finetuning #1 | 2.79% | 1.35% | 2.21% | 2.97% | 3.37% |
| Finetuning #2 | 2.43% | 2.18% | 0.93% | 2.90% | 3.18% |
| Finetuning #3 | 2.61% | 2.80% | 3.75% | 0.98% | 1.81% |
| Finetuning #4 | 3.12% | 3.51% | 2.72% | 1.63% | 0.78% |
Sharing content requires targeting cookies to be enabled. Please update your cookie preferences to use this feature.