INDEX

Explanations

obscurity and rarity

This neuron fires on tokens that are part of a dictionary-style word list with frequency or count annotations—that is, when the text is enumerating words alongside their numeric occurrence counts.

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

uent

-0.81

hap

-0.81

čních

-0.80

oidal

-0.77

apati

-0.76

byes

-0.75

šné

-0.74

betic

-0.74

vedev

-0.74

!!}

-0.74

POSITIVE LOGITS

 obscure

4.50

obscure

3.78

 rare

2.78

 obscurity

2.63

 uncommon

2.61

 niche

2.58

 esoteric

2.36

 rarer

2.28

rare

2.27

 unusual

2.20

Activations Density 0.180%