INDEX

Explanations

CWE vulnerability descriptions

The neuron strongly activates on uppercase metadata labels and markers (e.g. TEMPLATE, CWE identifiers, section headings) in code/comment headers.

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

人の

-1.36

resiasi

-1.27

in

-1.23

μών

-1.23

!”

-1.23

влении

-1.23

This

-1.21

 jika

-1.16

 Diskussion

-1.13

 incluir

-1.12

POSITIVE LOGITS

 сидит

1.57

 сына

1.54

糴

1.52

遄

1.48

眎

1.40

 testamento

1.38

 sweter

1.34

 Technologien

1.34

 paroisse

1.32

 genio

1.30

Activations Density 0.096%