Our new X account is live! Follow @wizwand_team for updates

Faithfulness evaluation on TellMeWhy

0.368AUC π-Soft-NS

Grad-ELLM

Updated 2d ago

Evaluation Results

Method	Links
Grad-ELLM 2026.01		0.368	0.531
Grad-ELLM 2026.01		0.368	0.531
Random 2026.01		0.335	0.564
Random 2026.01		0.335	0.564
DeepLIFT 2026.01		0.313	1.034
Saliency 2026.01		0.309	0.796
Grad-ELLM 2026.01		0.308	1.322
Input×Gradients 2026.01		0.308	0.462
Layer Gradient×Activation 2026.01		0.308	0.46
Input×Gradients 2026.01		0.308	0.462
Layer Gradient×Activation 2026.01		0.308	0.46
DeepLIFT 2026.01		0.307	0.467
DeepLIFT 2026.01		0.307	0.467
Saliency 2026.01		0.304	0.465
Saliency 2026.01		0.304	0.465
Input×Gradients 2026.01		0.299	0.745
Layer Gradient×Activation 2026.01		0.299	0.758
Integrated Gradients 2026.01		0.286	1.356
Random 2026.01		0.283	0.796
Integrated Gradients 2026.01		0.257	0.386
Integrated Gradients 2026.01		0.257	0.386
Attention 2026.01		0.247	0.426
Attention 2026.01		0.247	0.426
Attention 2026.01		0.225	0.684
Value Zeroing 2026.01		0.19	0.389
Value Zeroing 2026.01		0.19	0.389
Value Zeroing 2026.01		0.162	0.398