INDEX

Explanations

HTTPS POST calls

The neuron strongly activates on list‐style markers—especially colons and numerals—i.e. tokens that introduce or enumerate items.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

我们的

-1.16

 européennes

-1.13

 sœurs

-1.12

 célèbres

-1.10

świ

-1.09

 pierres

-1.08

 שלנו

-1.07

当前

-1.07

・・・・・

-1.04

 Redacción

-1.03

POSITIVE LOGITS

 both

1.30

😓

1.16

😢

1.13

◦

1.10

®

1.07

 latter

1.07

😥

1.07

xxi

1.06

xii

1.05

xD

1.04

Activations Density 0.149%