INDEX

Explanations

have to

The neuron strongly activates on informal auxiliary and possessive verbs—especially “got” and “have” (and their contractions like “you’ve” or “you’re got to”) in conversational phrases.

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

 didnt

0.75

 เคย

0.65

 applicator

0.64

 wrote

0.62

 fece

0.62

 hatte

0.60

 wasnt

0.59

етки

0.58

 machte

0.57

/'+

0.57

POSITIVE LOGITS

 phải

1.55

 must

1.52

必须

1.45

必须要

1.42

ต้อง

1.39

어야

1.39

must

1.38

 трябва

1.34

 Must

1.32

 باید

1.32

Activations Density 0.177%