INDEX

Explanations

words ending in 'l'

The neuron strongly activates on uncommon multi‐letter chunks—like proper names, acronyms, technical terms, or otherwise rare/long letter sequences—regardless of context.

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

↵↵

-3.13

lapsingToolbar

-2.50

みたいに

-2.48

 нього

-2.36

 worrying

-2.36

馀

-2.30



-2.28



-2.28

 autres

-2.16

 childish

-2.16

POSITIVE LOGITS

л

3.61

𓆜

3.38

琑

3.17

“

3.03

ﮨ

2.98

鹋

2.98

蛧

2.97

2.92

istically

2.91

Activations Density 1.547%