INDEX

Explanations

elements related to subscriptions and user engagement

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

news archives and access

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

centiles

-0.42

eseorang

-0.40

 duración

-0.38

mejores

-0.38

 Wachs

-0.36

phazard

-0.36

 saveiro

-0.36

ponible

-0.35

 nakalista

-0.35

 organizada

-0.35

POSITIVE LOGITS

 newspaper

0.63

Newspaper

0.63

 Newspaper

0.59

 journalistic

0.59

 journalism

0.58

 Newspapers

0.56

 journalists

0.54

 newspapers

0.52

twimg

0.52

 News

0.52

Activations Density 0.083%

elements related to subscriptions and user engagement

news archives and access

No Comments

No Known Activations