INDEX

Explanations

citation markers or code identifiers

The neuron fires on the “.” tokens used as property‐access operators in code (i.e. member-access dots).

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

 spéciale

-1.24

 május

-1.18

bigr

-1.16

ꪻ

-1.16

の方も

-1.15

pirable

-1.13

 индика

-1.12

ragend

-1.12

どれだけ

-1.11

 другие

-1.10

POSITIVE LOGITS

tion

1.15

もある

1.12

鈥

1.05

 gänzlich

1.03

—–

1.03

li

1.03

 geleg

1.02

 obicei

1.02

eneral

1.01

‒

1.01

Activations Density 0.003%