INDEX

Explanations

Please note or ignore

The neuron strongly fires on the little “helper” words and markers of an explanatory answer—things like “A:”, “Also,” “You will…,” “Let… be…,” and repeated conjunctions (“and,” “be,” “will”) that organize a solution or directive.

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

ош

-0.81

と呼ば

-0.81

ރ

-0.76

❄️

-0.74

ປ

-0.74

гай

-0.72

aky

-0.71

мере

-0.71

 consultez

-0.71

 kame

-0.71

POSITIVE LOGITS

 note

7.34

 Note

6.00

Note

5.88

note

5.56

 notice

4.47

 NOTE

3.98

NOTE

3.75

 notes

3.55

 Notice

3.53

 noted

3.52

Activations Density 0.079%