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

Input-based feature description faithfulness on GPT2 Res. SAE

60.4Faithfulness Score

EnsembleR (All)

22.23232.14142.0551.959Jan 14, 2025
Updated 4d ago

Evaluation Results

MethodLinks
2025.01
60.4
2025.01
59.6
2025.01
58.8
2025.01
44.4
2025.01
42.4
2025.01
29.2
2025.01
25.4
2025.01
23.7