INDEX

Explanations

often controversial

This neuron fires strongly on the very first substantive content word or phrase right after a segment boundary.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

 Ecco

0.90

 Oleh

0.85

 Hoon

0.82

 HMAC

0.82

Ecco

0.82

عيه

0.80

 jeudi

0.80

 OCLC

0.79

 Raha

0.78

 Mbps

0.78

POSITIVE LOGITS

Ι

0.91

蔔

0.77

美味しい

0.74

ändern

0.73

融入

0.71

ර්ධ

0.65

レンジ

0.64

 насыщен

0.64

共に

0.63

くだ

0.62

Activations Density 0.001%