INDEX

Explanations

code keywords and syntax

The neuron fires most strongly on core code‐syntax tokens—i.e. string delimiters (quotes, backticks), control-flow keywords (catch, return), and other punctuation that marks code structure.

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

 such

-1.75

 then

-1.63

 polícia

-1.28

オーナー

-1.23

 brinda

-1.23

같

-1.22

such

-1.20

 необходи

-1.16

Schule

-1.16

 nación

-1.16

POSITIVE LOGITS

 vierge

1.30

谙

1.27

yaret

1.20

 москве

1.19

 EVERYONE

1.16

 дороге

1.14

ALL

1.13

ybė

1.13

лива

1.10

taminophen

1.09

Activations Density 0.043%