INDEX

Explanations

code delimiters and keywords

The neuron strongly activates on closing‐brace tokens (the “}” that ends functions or blocks) in code.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

 that

-1.34

 with

-1.28

 then

-1.27

for

-1.24

 when

-1.23

 those

-1.14

by

-1.13

 after

-1.13

at

-1.11

 which

-1.10

POSITIVE LOGITS

 arrêter

1.20

 commencer

1.09

 prouver

1.03

 diep

1.02

AKA

1.02

 manier

1.01

peut

1.01



1.00

Шо

1.00

 Diſ

0.99

Activations Density 0.002%