INDEX

Explanations

years and Q:

The neuron strongly activates on numeric tokens—especially individual digits in multi‐digit years (e.g. the “9” in “1967”).

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

 Cruces

-1.17

 ако

-1.16

плю

-1.13

 jät

-1.10

 pás

-1.09

 Poni

-1.08

αυ

-1.07

 menik

-1.07

 Komun

-1.07

 kommunik

-1.06

POSITIVE LOGITS

1.31

1.24

ell

1.18

ah

1.16

les

1.14

ans

1.13

1.12

ys

1.10

le

1.09

ul

1.09

Activations Density 0.006%