Multi-object semantic discovery on CLEVR6
79.3mAccDTI-Sprites
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DTI-SpritesSprite generation=Pixels, Decision function=Min-Loss, Activation function=S, Training criterion=L0−1=comp, Regularization=Lempty, Training schedule=Longer training (†)2026.04 | 79.3 | 64.2 | |
| Ours-DSprite generation=MLP, Decision function=linear mapping, Activation function=GS, Training criterion=Lcomp, Regularization=Lempty2026.04 | 74.7 | 57.8 | |
| Ours-DSprite generation=MLP, Decision function=linear mapping, Activation function=GS, Training criterion=Lcomp, Regularization=Lempty + Lfreq2026.04 | 72.2 | 54.8 | |
| Ours-DSprite generation=MLP, Decision function=linear mapping, Activation function=GS, Training criterion=Lcomp, Regularization=None2026.04 | 66.8 | 49.6 | |
| Ours-DSprite generation=MLP, Decision function=linear mapping, Activation function=GS, Training criterion=Lcomp, Regularization=Lempty + Lbin2026.04 | 65 | 46.5 | |
| Ours-DSprite generation=MLP, Decision function=linear mapping, Activation function=GS, Training criterion=Lcomp, Regularization=τ2026.04 | 56.3 | 39.6 |