INDEX

Explanations

requires, must, need

The neuron detects uncommon, low‐frequency content tokens—i.e. rare or domain‐specific 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

if

-1.33

 this

-1.32

to

-1.26

することで

-1.09

 اگر

-1.06

 если

-1.04

这是

-1.03

can

-1.02

それでも

-1.02

これで

-0.99

POSITIVE LOGITS

 requires

2.48

requires

2.13

 Requires

1.96

Requires

1.85

 require

1.66

 requiere

1.63

 must

1.55

 need

1.48

 first

1.42

首先

1.42

Activations Density 0.147%