INDEX

Explanations

assigning expression as

The neuron strongly activates on very short tokens—single characters (like “/” or “a” or `"`) and brief function words (e.g. “as,” “for”).

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

 still

-1.31

of

-1.28

參考資料

-1.27

 createSprite

-1.23

 glyphicon

-1.22

service

-1.21

という感じ

-1.18

UCIÓN

-1.17

參考文獻

-1.16

眺望

-1.16

POSITIVE LOGITS

為に

1.95

割と

1.44

こともある

1.43

ほら

1.42

に追加

1.37

渽

1.37

えっ

1.36

様な

1.35

 plazos

1.29

ことも

1.26

Activations Density 0.009%