INDEX

Explanations

The neuron fires on verbs used in future‐tense or “will” constructions (i.e. action words that follow “will”)

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 crappy

0.48

 bisschen

0.47

 weird

0.45

 wasn

0.45

 horrible

0.45

ってる

0.45

 udah

0.45

 라고

0.45

してる

0.44

 avevo

0.44

POSITIVE LOGITS

 undertake

0.64

 become

0.55

 participate

0.55

become

0.54

采取

0.53

 pursue

0.52

 undergo

0.52

 receive

0.51

成为

0.50

 engage

0.50

Activations Density 2.105%