INDEX

Explanations

objective or purpose

This neuron detects section-heading tokens (especially “Objective(s)”, “Objectives”, “Purpose”, etc.) marking the start of a paper’s objective or purpose statement.

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

蓠

-1.89

を作り

-1.66

 detayl

-1.64

 relaciona

-1.55

 Eigentü

-1.54

ョウ

-1.53

 autén

-1.51

 consideradas

-1.51

馬上

-1.50

られていた

-1.50

POSITIVE LOGITS

for

2.02

 also

1.62

 również

1.61

and

1.57

白衣

1.54

."

1.53

This

1.50

="

1.48

𓋼

1.45

Activations Density 0.038%