INDEX

Explanations

punctuation and numerical formats

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

bibliography markers

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

 island

-0.29

 falle

-0.27

FXML

-0.27

en

-0.27

 Island

-0.26

angliski

-0.26

SceneManagement

-0.25

 caráter

-0.25

нъ

-0.25

 itself

-0.24

POSITIVE LOGITS

Diwedd

0.79

 pinulongan

0.65

 autorytatywna

0.65

yyr

0.64

mybatisplus

0.63

textra

0.63

AnimationsModule

0.63

 TextAppearance

0.62

rfan

0.62

peper

0.61

Activations Density 0.000%

punctuation and numerical formats

bibliography markers

No Comments

No Known Activations