List your company

ResearchFormerly WizWand

SAE Feature Interpretability

Benchmarks

DatasetSOTA methodMetricTrendResultsLast updated

Gemma 3 270M SAE Features (Value/Output)

21.82Input Interpretability Score

9

Jun 9, 2026

Qwen-3-1.7B SAE Features Value Output

6.96Interpretability Score

2

Jun 9, 2026

Qwen-3-1.7B SAE Features Key/Input

4.76Interpretability Score

2

Jun 9, 2026

Gemma SAE Features Value Output 3-1B

5.3Interpretability Score

2

Jun 9, 2026

Gemma SAE Features Key Input 3-1B

4.86Interpretability Score

2

Jun 9, 2026

Gemma-3-270M SAE Features Key Input

5.44Interpretability Score

2

Jun 9, 2026

GPT-2 Small SAE Features Value Output

5.95Interpretability Score

2

Jun 9, 2026

GPT-2 Small SAE Features Key Input

4.02Interpretability Score

2

Jun 9, 2026