INDEX

Explanations

code structure and legal terms

The neuron strongly activates on markup‐style tokens (e.g. HTML/XML tags like `<tr>`, `</th>`, `<td>` and similar code‐markup elements).

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

-1.52

-1.29

out

-1.15

of

-1.14

an

-1.12

 even

-1.10

new

-1.09

do

-1.08

 first

-1.07

one

-1.06

POSITIVE LOGITS

<h4>

1.70

<th>

1.67

acrylate

1.26

 ####

1.16

']['

1.16

/***

1.15

<<<<<<<

1.14

 ControllerBase

1.14

 uitgenodigd

1.14

 スーツ

1.13

Activations Density 0.000%