INDEX

Explanations

predicting next token in sequence

stories and descriptions

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

ρική

0.49

ра

0.46

 boutiques

0.45

નિક

0.43

tyard

0.43

rijven

0.43

না

0.42

ož

0.42

joke

0.42

 Dink

0.42

POSITIVE LOGITS

 güç

0.51

 apache

0.48

 sejam

0.47

 Invoke

0.46

 nici

0.46

apache

0.46

 Therapy

0.44

 Specify

0.44

 Whether

0.44

一台

0.44

Activations Density 0.002%