INDEX

Explanations

FitnessPal, Ameritrade, Ampere

The neuron strongly activates on subword tokens that are parts of capitalized names, titles, or acronyms (i.e. proper nouns and initials).

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

 Ката

0.25

rawd

0.25

 Entretanto

0.25

 Universit

0.24

 Temperatura

0.24

 Меди

0.24

 Институт

0.23

れる

0.22

RESS

0.22

 QuantityOnHand

0.22

POSITIVE LOGITS

 playbook

0.25

 launches

0.24

 của

0.23

 countries

0.23

 inflict

0.22

 capabilities

0.22

 systems

0.22

 facilities

0.22

 spearheaded

0.22

我现在

0.21

Activations Density 0.000%