INDEX

Explanations

citations like [cite] or [reference]

The neuron fires on non-content formatting and markup tokens—e.g. code-comment markers (//=), citation brackets ([@), and stray section/heading fragments (like the Greek “θέ”).

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

 mecánica

-2.67

廡

-2.64

 haupts

-2.61

形で

-2.52

簀

-2.47

 bijna

-2.44

⃨

-2.44

繾

-2.41

abz

-2.30

禵

-2.30

POSITIVE LOGITS

</strong>

3.38

 для

2.78

焄

2.73

2.64

 после

2.61

</th>

2.53

</h1>

2.48

 anderen

2.44

</sub>

2.41

 入れ

2.39

Activations Density 0.008%