INDEX

Explanations

unexpected outcome revealed

The neuron detects high-frequency auxiliary verbs (especially past‐tense and perfect forms like “was,” “were,” and “has”).

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

ها

-1.11

 あん

-1.10

 всегда

-1.08

法

-1.06

 oogen

-1.04

 sombrero

-1.01

 amigurumi

-1.01

自身

-1.01

 these

-1.01

SO

-1.01

POSITIVE LOGITS

 even

1.30

 prazo

1.28

 sogar

1.26

 refiri

1.23

 there

1.23

〝

1.17

 nawet

1.17

même

1.16

 dzięki

1.15

に行ってきました

1.14

Activations Density 0.112%