INDEX

Explanations

JMSSender JMSReceiver

The neuron strongly activates on code‐style identifiers and acronyms—i.e. tokens composed of uppercase letters or mixed‐case technical names.

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

 with

-1.41

at

-1.27

one

-1.25

on

-1.23

in

-1.21

 there

-1.16

 even

-1.14

all

-1.09

 through

-1.05

 only

-1.03

POSITIVE LOGITS

 предполага

1.08

 evtl

1.05

 normes

1.04

 myriad

1.03

 diverses

1.03

粝

1.02

beginning

1.01

 emplois

1.00

≒

1.00

砷

1.00

Activations Density 0.093%