INDEX

Explanations

obsessed with

The neuron flags occurrences of the root “obsess” (e.g. “obsessed,” “obsession”)—typically in “obsessed with” or “obsession with” constructions.

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

"));

-2.06

-1.98

༷

-1.90

-1.74

}}$

-1.74

</i>

-1.73

}}$

-1.69

the

-1.60

 doce

-1.60

 причиной

-1.59

POSITIVE LOGITS

⢌

1.56

↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵

1.51

蘼

1.48

 obtain

1.48

 ซึ่ง

1.48

↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵

1.47

 Ausblick

1.43

oudoune

1.43

ülete

1.41

 inhuman

1.40

Activations Density 0.006%