INDEX

Explanations

comment prefix or label

This neuron detects structural document markers and headings—tokens that denote section breaks or labels (like “Comment:”, “***”, “—–”, “* v0.8.2”, “[…]”, “Description:”, “Methods:”).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 cérami

-1.48

 résine

-1.44

 afecto

-1.41

 botanique

-1.38

 oameni

-1.37

 vaisselle

-1.36

 decidieron

-1.31

 trabajan

-1.31

 chré

-1.30

 negociar

-1.29

POSITIVE LOGITS

\\

1.41

 наши

1.40

Our

1.40

for

1.39

ON

1.29

 centavos

1.27

лама

1.27

her

1.23

 multitude

1.23

Activations Density 0.153%