INDEX

Explanations

speaker or interviewer

The neuron is a detector for numeric tokens (digits, line/timestamp numbers, speaker‐ID numbers, etc.).

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

aguya

-0.82

ButtonModule

-0.82

 viết

-0.82

CardModule

-0.79

으며

-0.77

 Dietrich

-0.77

 persoalan

-0.74

overe

-0.72

bedaan

-0.71

 Vikipedi

-0.71

POSITIVE LOGITS

 speaker

1.23

inaudible

1.11

 Speaker

1.06

 transcript

1.01

 speakers

1.00

Interviewer

0.99

speaker

0.96

Speaker

0.93

 interviewer

0.93

 Interview

0.92

Activations Density 0.005%