INDEX

Explanations

This neuron picks out words that signal the speaker’s stance or attitude—especially modals and sentiment markers (e.g. “would,” “have,” “care,” “really,” “sympathy”).

if/but clauses

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

 liksom

-0.94

 alltid

-0.94

 člověk

-0.89

 hvert

-0.88

kozások

-0.86

 encantó

-0.86

 både

-0.85

 familières

-0.84

 etmeye

-0.84

ったり

-0.83

POSITIVE LOGITS

if

3.03

but

2.03

 nếu

1.88

 если

1.75

 unless

1.68

 until

1.60

If

1.60

 except

1.59

if

1.36

ถ้า

1.35

Activations Density 0.031%