INDEX

Explanations

Columbus

The neuron is highly sensitive to numeric tokens—sequences of digits like phone numbers, postal codes, years, etc.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

⤴

-0.88

راً

-0.74

 końcu

-0.72

 hvil

-0.71

naire

-0.71

 McLaughlin

-0.71

DBHelper

-0.69

 coronary

-0.69

TLV

-0.68

SerializeField

-0.67

POSITIVE LOGITS

 Columbus

1.06

Columbus

0.87

万美元

0.73

 Lettres

0.72

nott

0.71

batik

0.70

 Clouds

0.69

 Turin

0.68

झ

0.68

 HTMLElement

0.68

Activations Density 0.015%