INDEX

Explanations

the presence of placeholders or indicators of new sections in text

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-2B @ 17-gemmascope-res-65k

Configuration

google/gemma-scope-2b-pt-res/layer_17/width_65k/average_l0_125

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

65,536

Data Type

float32

Hook Name

blocks.17.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



-0.87

__(/*!

-0.66

providedIn

-0.62

ViewInit

-0.59

 supposed

-0.57

первых

-0.57

^^^^^^^^

-0.56

 hereby

-0.56

globalData

-0.55

%>%

-0.54

POSITIVE LOGITS

 estekak

0.82

CloseOperation

0.77

AccessFile

0.65

FormTagHelper

0.60

 للمعارف

0.60

principalTable

0.59

Glück

0.57

oredCriteria

0.57

şiv

0.57

Composable

0.57

Activations Density 0.034%

the presence of placeholders or indicators of new sections in text

No Comments

No Known Activations