INDEX

Explanations

No Explanations Found

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-2B @ 18-gemmascope-mlp-65k

Configuration

google/gemma-scope-2b-pt-mlp/layer_18/width_65k/average_l0_88

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

65,536

Data Type

float32

Hook Name

blocks.18.hook_mlp_out

Hook Layer

Architecture

jumprelu

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

For

-0.49

унд

-0.45

 ای

-0.45

co

-0.45

 kohta

-0.44

 []),

-0.43

pre

-0.43

momile

-0.43

cheme

-0.42

ۇ

-0.42

POSITIVE LOGITS

OGND

0.88

Personensuche

0.76

yntaxException

0.68

 Мексичка

0.66

 AssemblyCompany

0.65

UnusedPrivate

0.63

GOTREF

0.62

Geplaatst

0.58

########.

0.56

Activations Density 0.000%

No Known Activations

This feature has no known activations.