INDEX

Explanations

accurate

This neuron responds to uppercase tokens—especially acronyms or all-caps names (e.g. stock tickers or capitalized proper-noun abbreviations).

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

 Surprise

0.51

菡

0.49

 Very

0.48

ยน

0.48

quear

0.47

 simplement

0.46

 Essentially

0.46

 superflu

0.46

昇

0.46

 gild

0.46

POSITIVE LOGITS

Ac

0.78

Acc

0.70

ademia

0.70

 akka

0.64

 Acosta

0.64

Ak

0.63

royd

0.63

Ack

0.62

Ac

0.62

 Acme

0.61

Activations Density 0.225%