INDEX

Explanations

code delimiters and structure

The neuron strongly activates on code identifier tokens—i.e. names of functions, routes, components, etc., typically multi‐word IDs joined by underscores or dashes.

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

 dieselbe

-1.16

 nombreuses

-1.13

 suscit

-1.10

 accompagner

-1.07

ěř

-1.07

 Avec

-1.04

 była

-1.04

 cresce

-1.03

 sämtliche

-1.02

 déterminer

-1.02

POSITIVE LOGITS

in

1.42

 with

1.41

 without

1.38

and

1.30

new

1.26

 from

1.17

but

1.16

now

1.16

1.15

so

1.15

Activations Density 0.006%