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

Lst

-0.39

IsValid

-0.39

sikker

-0.39

 contente

-0.36

 Barb

-0.36

Rey

-0.36

Pri

-0.35

 Estatal

-0.35

ële

-0.35

Barb

-0.35

POSITIVE LOGITS

DrawerToggle

5.00

WithIOException

0.68

 Numerade

0.65

TagMode

0.60

 ModelExpression

0.59

تقاوى

0.59

Jeografia

0.58

󠁴

0.57

 <=",

0.56

Personensuche

0.54

Activations Density 0.000%

No Known Activations

This feature has no known activations.