INDEX

Explanations

phrases or terms associated with categorization and classification

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

Francia

-0.24

 intérêt

-0.22

 especialidad

-0.22

 descar

-0.21

🖨

-0.21

 llamaba

-0.21

 Española

-0.21

 occasionally

-0.21

 cooperación

-0.20

Links

-0.20

POSITIVE LOGITS

 للاسماء

0.83

ValueStyle

0.81

 AssemblyCulture

0.79

 Wiktionnaire

0.79

 Paglinawan

0.73

 ویکی‌پدی

0.73

AndEndTag

0.71

 سكانية

0.70

səhifə

0.70

enumii

0.70

Activations Density 0.075%

phrases or terms associated with categorization and classification

No Comments

No Known Activations