INDEX

Explanations

phrases indicating authorship or attribution

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

from sender or editor

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

AddTagHelper

-0.46

 przecież

-0.45

neſs

-0.44

 canst

-0.44

tidaknya

-0.43

ußt

-0.42

ſelves

-0.42

 Chwiliwch

-0.40

otheek

-0.39

 læng

-0.39

POSITIVE LOGITS

 AssemblyCompany

0.68

enumi

0.60

帖最后由

0.59

 Frank

0.52

 AssemblyProduct

0.50

enumii

0.50

labelledby

0.50

 المعيارى

0.49

 Thorn

0.48

 horn

0.48

Activations Density 0.007%

phrases indicating authorship or attribution

from sender or editor

No Comments

No Known Activations