INDEX

Explanations

completed or significant findings

The neuron activates on dry, “report-style” discourse words—section‐heading or procedural markers (e.g. methods, general, modified, attempted, review, share)—that structure formal or technical text.

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

نین

-1.23

࣭

-1.22

 réponses

-1.16

 merve

-1.16

Judas

-1.16

adakan

-1.14

⼆

-1.12

Elektron

-1.09

スムーズ

-1.09

 paysages

-1.09

POSITIVE LOGITS

You

1.24

LY

1.20

是

1.20

1.17

</>

1.14

this

1.13

We

1.11

↵↵↵

1.10

 είναι

1.10

AS

1.08

Activations Density 0.242%