INDEX

Explanations

numerical values and their formatting

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

IFrame

Link

Negative Logits

 ſch

-0.52

 ſche

-0.47

 ſta

-0.47

ſelves

-0.47

 confezione

-0.46

 pioggia

-0.45

ſelf

-0.44

 enquête

-0.43

 inspección

-0.42

存于互联网档案馆

-0.42

POSITIVE LOGITS

ftagPool

0.57

ConstraintMaker

0.48

 ***!

0.48

 ffilmiau

0.43

Välislingid

0.42

GEBURTS

0.42

yntaxException

0.41

TokenNameLBRACE

0.40

()}>

0.39

AnimationsModule

0.39

Activations Density 0.005%

numerical values and their formatting

No Known Activations