INDEX

Explanations

that particular

The neuron is essentially a function‐word detector, firing most strongly on high-frequency grammatical words—especially the complementizer “that” and the copula “is.”

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

how

-1.82

why

-1.52

 first

-1.23

 when

-1.20

 poszczegól

-1.16

 cómo

-1.16

-1.13

 exactly

-1.12

 względem

-1.10

irsten

-1.05

POSITIVE LOGITS

ivably

1.42

quilibrium

1.34

нологі

1.32

鹀

1.32

,{

1.29

tigste

1.28

attro

1.27

covery

1.26

jiet

1.23

-\\

1.23

Activations Density 0.178%