INDEX

Explanations

code parsing and manipulation

The neuron chiefly fires on numeric literals (integers or floating‐point numbers) in source 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

 will

-1.26

and

-1.23

 geschenk

-1.06

 remporté

-1.05

chalten

-1.04

 illustratie

-1.02

but

-1.02

 ktorí

-1.02

 kochen

-0.99

 heilig

-0.99

POSITIVE LOGITS

 „,

0.99

メガネ

0.96

corbic

0.96

才可以

0.94

 hafta

0.93

我们需要

0.91

 mauvaise

0.90

 BufferedWriter

0.88

 inneces

0.86

 aplicable

0.86

Activations Density 0.016%