INDEX

Explanations

names and specific follow-ons

The neuron fires on isolated chunks of URL‐ or code‐style tokens—i.e. alphanumeric ID segments and punctuation from links.

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

as

-2.72

on

-2.41

in

-2.19

 with

-2.17

for

-1.89

at

-1.84

or

-1.84

 without

-1.77

 through

-1.70

 when

-1.67

POSITIVE LOGITS

those

1.69

 kostenlose

1.68

 которы

1.68

mbal

1.62

 mejoras

1.61

킁

1.55

 rác

1.55

 ARISING

1.55

primarily

1.55

 ceux

1.55

Activations Density 0.000%