INDEX

Explanations

name prefixes or standalone names

The neuron fires primarily on function or constructor identifiers and their chaining punctuation (e.g. “Promise”, the “.” operator, “g” in code/math), i.e. it detects function‐name tokens and the dot accessor.

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

 both

-1.69

 their

-1.57

 only

-1.31

by

-1.30

it

-1.30

its

-1.29

 provide

-1.25

 swiftly

-1.25

 more

-1.22

 meticulous

-1.20

POSITIVE LOGITS

usually

1.42

 также

1.28

seems

1.24

often

1.23

 horrid

1.20

mainly

1.19

particularly

1.17

generally

1.16

knew

1.16

 Эти

1.15

Activations Density 0.000%