INDEX

Explanations

proper nouns, particularly names of authors and contributors

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

Chinese names and code contexts

np_acts-logits-general · gemini-2.5-flash-lite

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 IBOutlet

-0.47

 gòn

-0.41

 vent

-0.41

govina

-0.39

╤

-0.38

 spoiling

-0.37

Роз

-0.37

 spoil

-0.37

Pl

-0.36

ǫ

-0.35

POSITIVE LOGITS

 AssemblyCompany

0.62

webElementXpaths

0.59

ValueStyle

0.56

:✨

0.52

GEBURTSDATUM

0.51

 surla

0.50

basicConfig

0.50

endpush

0.50

تقاوى

0.49

Šaltiniai

0.47

Activations Density 0.001%

proper nouns, particularly names of authors and contributors

Chinese names and code contexts

No Comments

No Known Activations