INDEX

Explanations

terms indicating the highest quality or superlative status

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

 ModelExpression

-0.50

 gruesa

-0.47

 réfrig

-0.47

 فريبيس

-0.45

 bénéficiaire

-0.44

 éclairage

-0.41

RenderAtEndOf

-0.41

 duquel

-0.40

 Moscú

-0.40

 coû

-0.39

POSITIVE LOGITS

 quintessential

0.77

 perfection

0.65

 classic

0.60

 coolest

0.59

 epitome

0.59

 perfect

0.58

 ultimate

0.57

 cool

0.55

xase

0.55

 epitom

0.55

Activations Density 0.021%

terms indicating the highest quality or superlative status

No Comments

No Known Activations