INDEX

Explanations

nouns indicating significant concepts or entities within a 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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 quick

-0.44

few

-0.41

↵

-0.41

 close

-0.40

 less

-0.39

 also

-0.38

 closer

-0.37

 slow

-0.37

 gentle

-0.36

 Asian

-0.36

POSITIVE LOGITS

entire

0.89

Entire

0.88

 Entire

0.80

whole

0.77

 ENTIRE

0.75

 entire

0.75

 keseluruhan

0.74

 inteiro

0.73

 seluruh

0.72

########.

0.71

Activations Density 0.047%

nouns indicating significant concepts or entities within a text

No Comments

No Known Activations