INDEX

Explanations

references to Australia and its variations in text

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

fromnode

-0.47

 }^{[

-0.44

posedge

-0.41

ViewFeatures

-0.41

 المرك

-0.41

formKey

-0.40

selben

-0.39

 meds

-0.38

fileprivate

-0.38

ljiv

-0.38

POSITIVE LOGITS

Australia

1.01

 Australia

0.97

Australian

0.90

 Australian

0.90

 Austr

0.88

 Australians

0.86

 Australien

0.82

 Austral

0.81

australia

0.79

 australia

0.79

Activations Density 0.006%

references to Australia and its variations in text

No Known Activations