INDEX

Explanations

No Explanations Found

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-9B-IT @ 20-gemmascope-res-131k

Configuration

google/gemma-scope-9b-it-res/layer_20/width_131k/average_l0_81

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

131,072

Data Type

float32

Hook Name

blocks.20.hook_resid_post

Hook Layer

Architecture

jumprelu

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

AsUp

-0.60

 createSlice

-0.56

ngdoc

-0.55

ngrx

-0.54

msgTypes

-0.52

 actionTypes

-0.50

ViewImports

-0.47

transQ

-0.47

AndroidJUnit

-0.47

 snippetHide

-0.46

POSITIVE LOGITS

ISupport

0.78

//

0.44

the

0.43

WithMany

0.38

 pernas

0.38

 pets

0.38

 vôtre

0.38

/*

0.38

 obrazov

0.37

oshiba

0.37

Activations Density 0.000%

No Known Activations

This feature has no known activations.