INDEX

Explanations

proper nouns, specifically names of individuals

oai_token-act-pair · gpt-4o-mini

Triggered by @bot

names preceded by by

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

Negative Logits

ⓧ

-0.56

➯

-0.55

RegressionTest

-0.52

 Infórmanos

-0.51

 increí

-0.50

Personensuche

-0.48

 lenker

-0.48

 виправивши

-0.47

SharedCtor

-0.47

MethodManager

-0.47

POSITIVE LOGITS

AutoScaleMode

0.39

المصادر

0.39

 Samuel

0.37

handleClose

0.37

wire

0.37

 gynhyrchwyd

0.37

 wire

0.35

BY

0.35

zegor

0.34

Gau

0.34

Activations Density 0.032%

proper nouns, specifically names of individuals

names preceded by by

No Known Activations