INDEX

Explanations

software systems and implementation

The neuron strongly responds to unusually long (multi-syllabic) tokens, essentially marking very long words.

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

 katastro

-1.00

breviations

-1.00

 Благодаря

-0.96

 Meksiko

-0.94

 eksklu

-0.94

かしい

-0.93

establishing

-0.91

diging

-0.90

kkor

-0.90

らかに

-0.90

POSITIVE LOGITS

 prototype

2.19

 system

1.87

 tool

1.84

prototype

1.64

 prototypes

1.61

 platform

1.50

 Prototype

1.43

Prototype

1.36

 developed

1.36

 software

1.36

Activations Density 0.098%