INDEX

Explanations

phrases and terms of address directed towards readers or audiences

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

addresses to groups

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

 []).

-0.45

ИА

-0.40

 Lightboxes

-0.40

isional

-0.39

 Савезне

-0.39

ruka

-0.39

 Fries

-0.39

 يتيمه

-0.38

fej

-0.37

 []),

-0.36

POSITIVE LOGITS

 CURIAM

0.49

 StatelessWidget

0.48

 pembaca

0.48

大家好

0.47

Климат

0.47

Sklici

0.46

 CreateTagHelper

0.46

 StatefulWidget

0.46

Демографія

0.44

SerializedSize

0.43

Activations Density 0.010%

phrases and terms of address directed towards readers or audiences

addresses to groups

No Comments

No Known Activations