INDEX

Explanations

in, with, to

This neuron detects words that indicate providing help, assistance, or explanatory guidance (e.g. help, aid, find, understanding).

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

都沒

-1.20

ां

-1.02

for

-0.98

るので

-0.97

の影響

-0.96

]<<

-0.96

日が

-0.95

 cuidadosamente

-0.95

ாம்

-0.93

それで

-0.91

POSITIVE LOGITS

 better

1.52

 بهتر

1.27

 easier

1.23

 understand

1.17

一些

1.14

 někter

1.13

 navigate

1.09

更加

1.03

 some

0.98

如果

0.98

Activations Density 0.040%