INDEX

Explanations

phrases or statements that imply contradiction or uncertainty

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

 betweenstory

-0.57

 ſta

-0.52

tagHelperRunner

-0.48

 Paglinawan

-0.48

 faſt

-0.48

ſelf

-0.47

 ſte

-0.47

 ſhe

-0.47

 deſt

-0.46

 ſhould

-0.45

POSITIVE LOGITS

0.67

the

0.66

 something

0.56

IsContent

0.55

—

0.50

either

0.50

something

0.50

เฉพาะ

0.48

 marrom

0.48

Activations Density 0.043%

phrases or statements that imply contradiction or uncertainty

No Comments

No Known Activations