INDEX

Explanations

Japanese greetings and responses

This neuron responds to non-Latin or “special” tokens—e.g. Japanese or Korean characters, East Asian punctuation marks, emojis, and hashtag symbols.

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

ଗ

1.07

ଯ

1.05

 ବି

1.05

 alleging

1.02

 Demonstrated

1.02

 Undoubtedly

1.01

କୁ

0.98

ଭ

0.95

ifference

0.95

abhave

0.95

POSITIVE LOGITS

ちなみに

1.33

この

1.27

また

1.23

※

1.23

これは

1.22

今回は

1.20

これ

1.18

お

1.13

まずは

1.12

<0xE3>

1.11

Activations Density 0.016%