INDEX

Explanations

places to visit

The neuron detects imperative or recommendation cues—action‐oriented verbs (like “go,” “park,” “walk,” “take”) that give directions or advice.

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

 pulau

-1.31

vención

-1.30

 ぬいぐるみ

-1.29

 jaja

-1.26

rée

-1.25

 bukit

-1.24

 acquisto

-1.21

 Dateien

-1.20

alucía

-1.20

jaj

-1.17

POSITIVE LOGITS

 where

1.24

 sowohl

1.20

ซึ่ง

1.20

 włosów

1.14

이며

1.13

 ricas

1.13

Both

1.09

if

1.07

 visit

1.06

こちらも

1.06

Activations Density 0.010%