INDEX

Explanations

technical and abstract concepts

The neuron activates on specialized or technical vocabulary—words that tend to appear in medical, engineering, or other highly domain-specific contexts.

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

あぁ

-1.43

すみません

-1.24

 umożli

-1.18

While

-1.16

 allowing

-1.12

lable

-1.09

んですよね

-1.08

଼

-1.08

»).

-1.07

総評

-1.06

POSITIVE LOGITS

なども

1.22

1.19

鼍

1.13

fono

1.11

ilih

1.07

BBA

1.06

鈥

1.05

camis

1.05

 ويل

1.04

 gallons

1.02

Activations Density 0.042%