INDEX

Explanations

named entities, particularly names of characters and places

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

New Auto-Interp

Configuration

google/gemma-scope-9b-pt-res/layer_20/width_131k/average_l0_114

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

 dourada

-0.55

 riuscito

-0.52

 econômica

-0.51

 irmã

-0.50

rillation

-0.48

 tatuagem

-0.48

 potuto

-0.47

 lettore

-0.47

 femininas

-0.47

 irmão

-0.46

POSITIVE LOGITS

 betweenstory

0.64

(«

0.54

Portail

0.53

LBS

0.46

Viited

0.46

 AssemblyCulture

0.45

BIS

0.43

échal

0.43

出版年

0.42

LW

0.39

Activations Density 0.172%

named entities, particularly names of characters and places

No Comments

No Known Activations