Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Output-based feature description faithfulness on Llama MLP SAE 3.1
Loading...
55.5
Faithfulness Score
EnsembleC (All)
45.412
48.031
50.65
53.269
Jan 14, 2025
Faithfulness Score
Updated 1mo ago
Evaluation Results
Method
Method
Links
Faithfulness Score
EnsembleC (All)
Layer Aggregation=Aver...
2025.01
55.5
TokenChange
Layer Aggregation=Aver...
2025.01
53.1
EnsembleR (All)
Layer Aggregation=Aver...
2025.01
51.6
EnsembleR (VP+TC)
Layer Aggregation=Aver...
2025.01
50.7
MaxAct
Layer Aggregation=Aver...
2025.01
49.6
EnsembleR (MA+TC)
Layer Aggregation=Aver...
2025.01
48.9
VocabProj
Layer Aggregation=Aver...
2025.01
48.2
EnsembleR (MA+VP)
Layer Aggregation=Aver...
2025.01
45.8
Feedback
Search any
task
Search any
task