Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Feature Interpretability on LLM Activations
Loading...
86.9
AutoInterp Score
HSAE
80.452
82.126
83.8
85.474
Feb 12, 2026
AutoInterp Score
SCR Score
Updated 1mo ago
Evaluation Results
Method
Method
Links
AutoInterp Score
SCR Score
HSAE
Dictionary Size=16384
2026.02
86.9
32.4
Baseline
Dictionary Size=16384
2026.02
86.1
33.8
HSAE
Dictionary Size=8192
2026.02
85.9
31.4
Baseline
Dictionary Size=8192
2026.02
85.4
32
Baseline
Dictionary Size=4096
2026.02
82.9
30.2
HSAE
Dictionary Size=4096
2026.02
81.4
28.6
Baseline
Dictionary Size=2048
2026.02
81.3
24.5
HSAE
Dictionary Size=2048
2026.02
80.7
27.2
Feedback
Search any
task
Search any
task
