INDEX

Explanations

references to Hinduism and related cultural identifiers

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

 Autorisations

-0.70

例句

-0.59

mittances

-0.54

 مشين

-0.53

 Terme

-0.53

 Autorizaciones

-0.53

elsa

-0.53

teren

-0.50

ebe

-0.50

oneta

-0.49

POSITIVE LOGITS

 Hindu

0.90

Hindu

0.82

 hindu

0.69

 Hindus

0.66

hindu

0.60

 Hinduism

0.55

INDU

0.51

 dieux

0.46

 senderismo

0.44

 setPassword

0.44

Activations Density 0.003%

references to Hinduism and related cultural identifiers

No Comments

No Known Activations