INDEX

Explanations

dialogue speaker names

This neuron fires on speaker labels (names followed by a colon) in dialogue transcripts.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

 माने

0.58

 rije

0.57

 yourselves

0.55

 spotless

0.55

 божомолу

0.55

परेट

0.54

 kalian

0.54

 অবস্থান

0.53

 पीडीएफ

0.51

nią

0.51

POSITIVE LOGITS

1.10

:...

0.97

：

0.94

0.91

:+

0.85

:${

0.84

:"

0.81

:(

0.81

:“

0.80

.:

0.79

Activations Density 0.007%