INDEX

Explanations

negative contractions

The neuron fires on semantically contentful or “topical” words—i.e. nouns, verbs, adjectives, and other informative tokens—rather than on common function words.

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

not

-1.33

 نه

-1.17

not

-1.12

又不是

-0.93

かもし

-0.84

らしいです

-0.83

 NOTA

-0.82

 هیچ

-0.80

 więks

-0.79

 neither

-0.78

POSITIVE LOGITS

isn

3.88

 doesn

3.78

 didn

3.75

 wouldn

3.59

 aren

3.52

don

3.28

 hasn

3.00

 wasn

2.98

Doesn

2.94

Didn

2.81

Activations Density 0.085%