INDEX

Explanations

Data, blog posts, mapping

This neuron detects LaTeX command tokens (especially math‐mode macros like “\text{” or “\mathbb{”) by firing on the literal “text{”, “mathbb{”, 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

</em>

-3.06

</h4>

-2.53

-2.19

“

-2.05

</td>

-2.05

${

-1.99

 för

-1.98

 whatnot

-1.93

-1.88

POSITIVE LOGITS

睒

2.00

 ".";

1.93

?).

1.88

 cabrio

1.84

imanapun

1.83

Ặ

1.83

…).

1.81

 camry

1.80



1.80

「

1.80

Activations Density 0.001%