INDEX

Explanations

references to news and media content

oai_token-act-pair · gpt-4o-mini Triggered by @bot

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

featureID

-0.49

ReusableCell

-0.48

 autorytatywna

-0.46

 Pardo

-0.42

providedIn

-0.42

 GenerationType

-0.41

 hObject

-0.41

рк

-0.40

IContainer

-0.40

cnico

-0.40

POSITIVE LOGITS

twimg

0.55

 journaux

0.54

 daily

0.45

 newsletters

0.44

 news

0.42

 sledo

0.42

 periodicals

0.41

 newspapers

0.40

 diariamente

0.40

 protoimpl

0.40

Activations Density 0.049%

references to news and media content

No Comments

No Known Activations