INDEX

Explanations

repeated references to titles and names in the context of entertainment and culture

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

names or titles of entities

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

 ModelExpression

-0.79

Personendaten

-0.79

 للاسماء

-0.70

 resourceCulture

-0.69

 kasarigan

-0.69

 CanadaChoose

-0.69

ValueStyle

-0.68

XmlAccessType

-0.68

出版年

-0.68

angliski

-0.67

POSITIVE LOGITS

 vinto

0.36

surate

0.36

heizung

0.36

zeile

0.36

 judiciaire

0.34

grès

0.33

AuthGuard

0.32

 tecnici

0.31

firmasi

0.31

nEnter

0.31

Activations Density 0.093%

repeated references to titles and names in the context of entertainment and culture

names or titles of entities

No Comments

No Known Activations