INDEX

Explanations

option lists with values

This neuron detects the structure of multiple-choice questions—especially the question prompt (“What is…”) and the labeled answer options (a), (b), (c), etc.

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

 покупке

-1.66

 descendre

-1.61

(!__

-1.61

爱心

-1.48

 kahve

-1.46

 natten

-1.45

疥

-1.43

żli

-1.41

 мероприятий

-1.41

aned

-1.39

POSITIVE LOGITS

 preservar

1.58

 While

1.50

 When

1.48

 with

1.45

to

1.45

 still

1.42

 what

1.42

 лишь

1.39

 Once

1.39

but

1.39

Activations Density 0.032%