INDEX

Explanations

code delimiters

The neuron fires on specific named‐entity strings—capitalized names, quoted phrases, file‐paths/URLs, code or tag snippets, and similar multi-token proper references.

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

 Although

1.05

 Однако

1.04

 Ironically

1.04

 When

1.01

But

1.00

 ซึ่ง

1.00

ซึ่ง

0.97

 During

0.96

 However

0.95

 ولكن

0.95

POSITIVE LOGITS

 altid

0.87

 purposeful

0.86

 spher

0.81

 propulsion

0.78

 myriad

0.77

 puedas

0.77

 immer

0.77

 spontaneous

0.76

 liberation

0.76

 unbounded

0.75

Activations Density 0.000%