INDEX

Explanations

issue followed by topic

The neuron fires strongly on occurrences of “issue” or “Issues” (often in question titles), i.e. when a problem-statement heading appears.

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

].

-2.83

 arbeitete

-2.77

는

-2.48

 highest

-2.45



-2.42

。

-2.33

 Vorsitzender

-2.27

 investimento

-2.27

および

-2.27

POSITIVE LOGITS

鈇

2.44

 dotyczące

2.28

 också

2.28

libatkan

2.06

 nytt

2.03

嗞

1.98

 magasiner

1.96

 continúas

1.95

 cirka

1.95

 året

1.94

Activations Density 0.028%