INDEX

Explanations

references to the term "man" or its variations

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

kilde

-0.48

yll

-0.46

orios

-0.45

 inmediato

-0.45

xiu

-0.44

TokenNameR

-0.41

uque

-0.41

reshold

-0.41

笈

-0.41

 Bildungs

-0.41

POSITIVE LOGITS

Man

1.20

Man

1.14

man

1.14

man

1.02

MAN

0.98

MAN

0.88

 Manning

0.82

 Mann

0.81

 Woman

0.79

 woman

0.74

Activations Density 0.013%

references to the term "man" or its variations

No Comments

No Known Activations