INDEX

Explanations

instances of the word "well" used to express a response or opinion

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

discourse marker well

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

脚注の使い方

-0.58

IsMutable

-0.54

MergeFrom

-0.50

SequentialGroup

-0.49

BibitemShut

-0.48

ьаж

-0.48

Atoi

-0.46

IntoConstraints

-0.45

SerializedSize

-0.45

Искәрмәләр

-0.45

POSITIVE LOGITS

 Савезне

0.46

 незавершена

0.43

verwijspagina

0.40

انجليز

0.38

 ModelExpression

0.36

 aika

0.36

 ujednoznacz

0.35

 للاسماء

0.34

sort

0.33

 betweenstory

0.33

Activations Density 0.017%

instances of the word "well" used to express a response or opinion

discourse marker well

No Comments

No Known Activations