INDEX

Explanations

terms related to inventions and their preferred embodiments

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



-0.81

цездатний

-0.73

RegressionTest

-0.65

FunctionFlags

-0.64

 Wikimedijinoj

-0.62

delwed

-0.60

Демографія

-0.59

 estekak

-0.59

 nakalista

-0.56

audiovisuel

-0.56

POSITIVE LOGITS

 idées

0.35

ngrx

0.31

 alike

0.31

 Idee

0.29

 nouvel

0.28

 propuesto

0.28

 للمعارف

0.28

 rowspan

0.28

,:)

0.28

 recettes

0.27

Activations Density 0.121%

terms related to inventions and their preferred embodiments

No Comments

No Known Activations