INDEX

Explanations

's

This neuron is highly active on words and punctuation that mark the start of new sentences or major segments (e.g. sentence-initial tokens and section headers).

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

 Sachin

0.27

 Although

0.27

 Firstly

0.27

 Özellikle

0.26

 Obtaining

0.26

The

0.25

लेकिन

0.25

 ஒரு

0.25

 Plaintiff

0.25

 Lorsqu

0.25

POSITIVE LOGITS

and

0.27

.'

0.26

in

0.24

0.23

on

0.23

 ashore

0.23

;'

0.22

,—

0.22

."

0.22

の

0.22

Activations Density 0.354%