INDEX

Explanations

making decisions

The neuron fires on single words that signal high importance or weight—terms like “progress,” “significant,” “conclusions,” “decisions,” or “damage.”

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

+}\

-1.95

FieldOffset

-1.76

*}[

-1.63

 meist

-1.59

 liebt

-1.52

二日

-1.47

 котором

-1.45

 nég

-1.45

that

-1.43

quelize

-1.43

POSITIVE LOGITS

in

2.63

on

2.11

for

1.98

 with

1.84

 against

1.67

 notare

1.65

 novità

1.53

などは

1.47

 dirit

1.46

 recensioni

1.45

Activations Density 0.272%