INDEX

Explanations

recurring phrases related to instructions or methods of doing tasks

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-9B-IT @ 20-gemmascope-res-131k

Configuration

google/gemma-scope-9b-it-res/layer_20/width_131k/average_l0_81

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

 brancas

-0.46

ftagPool

-0.45

发表于

-0.42

婿

-0.41

viewBox

-0.41

cillos

-0.39

 femininos

-0.39

 femininas

-0.39

ningarna

-0.38

NSDictionary

-0.38

POSITIVE LOGITS

 pinulongan

0.59

NameInMap

0.53



0.49

 autorytatywna

0.46

 /\.(

0.44

 Scotsman

0.44

 disambiguazione

0.43

ganu

0.42

Tikang

0.41

Portail

0.41

Activations Density 0.042%

recurring phrases related to instructions or methods of doing tasks

No Comments

No Known Activations