INDEX

Explanations

phrases related to deception or manipulation of information

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

appearing convincing or impressive

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

bereitung

-0.35

 jeunesse

-0.34

horabuena

-0.33

่าน

-0.33

Jahr

-0.33

 dúvidas

-0.32

 year

-0.32

 titulaire

-0.31

 illustration

-0.31

 Knight

-0.31

POSITIVE LOGITS

 ligiloj

0.59

 EconPapers

0.56

AccessorTable

0.56

ftagPool

0.55

Fake

0.54

CppMethod

0.54

tonode

0.53

 Fake

0.53

出版年

0.52

 fake

0.50

Activations Density 0.032%

phrases related to deception or manipulation of information

appearing convincing or impressive

No Comments

No Known Activations