INDEX

Explanations

attenuation #define tag name flight

The neuron detects unusually long, uninterrupted tokens (for example long words or code identifiers).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 peculiarity

-0.99

 carcinogenic

-0.95

CHAMP

-0.90

TERR

-0.88

inai

-0.87

PUBL

-0.86

 våra

-0.85

 fasti

-0.84

 travaillons

-0.84

czki

-0.82

POSITIVE LOGITS

mappedBy

1.15

Having

0.98

 reservar

0.96

 détru

0.96

How

0.95

 novelas

0.94

 milioane

0.93

getSystemService

0.93

Before

0.92

 tubos

0.91

Activations Density 0.138%