INDEX

Explanations

references to names or naming conventions in various contexts

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

name or person identifier

np_acts-logits-general · gemini-2.5-flash-lite

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

 &___

-0.61

ẵn

-0.48

хьтан

-0.47

jspb

-0.45

Læs

-0.45

vallis

-0.45

 estekak

-0.45

ọa

-0.44

+#+#

-0.44

 ivelany

-0.44

POSITIVE LOGITS

Personendaten

0.63

 person

0.56

 name

0.55

 Name

0.54

 getName

0.54

 NAME

0.51

姓名

0.50

 surname

0.50

 firstName

0.49

 perſon

0.48

Activations Density 0.006%

references to names or naming conventions in various contexts

name or person identifier

No Comments

No Known Activations