INDEX

Explanations

code block starters or separators

The neuron is highly sensitive to numeric literals (numbers) in 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

શ

-1.13

 revamped

-1.07

ellery

-1.06

 chcia

-1.05

 magnific

-1.03

 actes

-1.01

 berechnet

-1.01

 zwischen

-1.01

 também

-1.00

镲

-1.00

POSITIVE LOGITS

 with

1.47

 where

1.46

as

1.40

 after

1.39

 while

1.34

 still

1.32

on

1.25

1.23

 these

1.20

 using

1.20

Activations Density 0.013%