INDEX

Explanations

various forms of numeric or alphanumeric representations

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

surprise, surprise, electrical, polish, conclusion, consinstently, tru, leads

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

 незавершена

-0.71

IsMutable

-0.66

 Chwiliwch

-0.65

rungsseite

-0.64

yarnpkg

-0.63

 ModelExpression

-0.60

 emailAlready

-0.60

 disambiguazione

-0.59

ValueStyle

-0.57

 Geſch

-0.57

POSITIVE LOGITS

 statement

0.32

gründe

0.32

",{

0.31

underset

0.31

 STATEMENT

0.30

tfrac

0.30

 mock

0.29

Encoded

0.29

Statistic

0.29

 benzer

0.29

Activations Density 0.224%

various forms of numeric or alphanumeric representations

surprise, surprise, electrical, polish, conclusion, consinstently, tru, leads

No Comments

No Known Activations