INDEX

Explanations

section headers or code comments

This neuron responds strongly to non‐word, punctuation‐heavy tokens—especially hash signs (“#”) and signed numeric tokens (e.g. “-8”, “-16.75”) used in markup or attribute syntax.

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

↵↵↵↵

-1.69

↵↵↵

-1.58

↵↵↵↵↵↵↵

-1.52

↵↵↵↵↵

-1.51

rschein

-1.47

↵↵↵↵↵↵

-1.45

-1.41

```

-1.38

竄

-1.36

-1.35

POSITIVE LOGITS

 綺麗

1.57

 interessante

1.48

hålla

1.41

以外の

1.36

Să

1.34

 passende

1.34

izate

1.34

 peuvent

1.33

женские

1.33

bolsa

1.31

Activations Density 0.000%