INDEX

Explanations

names of people and titles

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

Chinese, Japanese, or French closings

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

 kasarigan

-0.53

yyr

-0.46

 einsch

-0.45

cbr

-0.43

chke

-0.42

förder

-0.41

tvguidetime

-0.41

 accession

-0.41

 tartalomajánló

-0.40

erm

-0.40

POSITIVE LOGITS

AutoScaleMode

0.57

 useAppContext

0.48

eseorang

0.47

 AssemblyCompany

0.43

\{\\

0.42

Cordialement

0.42

 jScrollPane

0.41

Tracce

0.40

DataSetChanged

0.39

thank

0.39

Activations Density 0.013%

names of people and titles

Chinese, Japanese, or French closings

No Comments

No Known Activations