INDEX

Explanations

statements or reports made by sources or insiders

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

reporting speech from sources

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

 surla

-0.63

 GenerationType

-0.60

abestanden

-0.59

OrNil

-0.56

 EconPapers

-0.56

featureID

-0.54

 дописавши

-0.53

 NSCoder

-0.53

astify

-0.52

Portale

-0.51

POSITIVE LOGITS

edor

0.40

Another

0.39

 another

0.39

 insider

0.39

eseorang

0.39

 anonymous

0.38

 Another

0.36

 anonymously

0.36

 חיצוניים

0.36

 secret

0.35

Activations Density 0.018%

statements or reports made by sources or insiders

reporting speech from sources

No Comments

No Known Activations