INDEX

Explanations

algorithms and techniques

The neuron fires strongly on multiword, capitalized technical method or algorithm names (e.g. “Block Preconditioned Conjugate Gradient Method”) and other prominent methodological terms.

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

即是

-1.04

Producción

-1.02

 tatuagens

-1.01

色々

-0.98

 sarebbe

-0.95

❾

-0.94

στά

-0.94

偁

-0.93

saia

-0.93

زارش

-0.93

POSITIVE LOGITS

 method

2.20

 technique

2.16

 approach

2.06

 with

1.74

 algorithm

1.64

 idea

1.38

for

1.35

 procedure

1.34

 framework

1.30

 techniques

1.23

Activations Density 0.149%