INDEX

Explanations

adjectives that denote high skill or positive attributes

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

descriptor followed by a noun

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

prnewswire

-0.40

 Sebab

-0.39

 அத

-0.38

Cuar

-0.38

więcej

-0.37

 idén

-0.37

 nahilalakip

-0.37

abz

-0.37

 Masalah

-0.36

 Question

-0.36

POSITIVE LOGITS

HasForeignKey

0.64

 fubject

0.59

 MainAxisSize

0.57

posedge

0.55

存于互联网档案馆

0.54

 мәкал

0.54

apsed

0.53

 poffe

0.52

PerformLayout

0.51

SharedCtor

0.51

Activations Density 0.044%

adjectives that denote high skill or positive attributes

descriptor followed by a noun

No Comments

No Known Activations