INDEX

Explanations

'where' introducing definitions

This neuron detects the “where” clauses (and related definitional words like “is” or “represents”) that introduce or explain symbols in technical or mathematical statements.

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

of

-2.13

in

-1.53

なども

-1.49

 dieses

-1.38

Approximate

-1.38

كويكب

-1.37

 tohoto

-1.34

 with

-1.31

装备

-1.30

 początku

-1.30

POSITIVE LOGITS

屨

1.55

 nôtre

1.41

práv

1.41

 ehemal

1.40



1.38

laus

1.35

'].'

1.34

ridae

1.34

epsilon

1.34

Activations Density 0.007%