INDEX

Explanations

random number, list, math functions

The neuron fires on programming‐related tokens—especially module/library names and import/assert keywords—i.e. it detects code import and API function identifiers.

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

 these

-1.42

 její

-1.08

Asts

-1.07

上記の

-1.05

 those

-1.05

Estos

-1.04

 mindre

-1.02

 ralenti

-1.00

当時は

-1.00

 namely

-1.00

POSITIVE LOGITS

гова

1.29



1.25

 diinginkan

1.25

追求

1.24

 berdua

1.16

1.15

ponenten

1.13

közi

1.12

ọi

1.10

⌇

1.10

Activations Density 0.016%