INDEX

Explanations

hyphenated words or abbreviations

The neuron is looking for tokens characteristic of a structured instructional document or tutorial (e.g. section headings and descriptive metadata in a Markdown‐style “Practical Python Programming” course outline).

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

}-${

-1.20

on

-1.09

seqs

-1.05

其他

-1.04

addrs

-1.03

 svært

-1.02

 eerst

-1.01

 gjør

-0.98

 vanaf

-0.97

 aforementioned

-0.96

POSITIVE LOGITS

сында

1.17

/}{

1.11

龛

1.09

ヤバ

1.02

 olabilir

1.00

ⓕ

1.00

Bjs

0.98

 vinho

0.98

的朋友

0.96

phòng

0.96

Activations Density 0.121%