INDEX

Explanations

hard

The neuron detects occurrences of the letter sequence “Hard” within words.

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

etan

-0.96

ead

-0.94

</em>

-0.92

 warmly

-0.86

 bleus

-0.84

 терпе

-0.84

experiment

-0.82

ecg

-0.81

ights

-0.80

 sliced

-0.80

POSITIVE LOGITS

 hard

1.80

Hard

1.73

 Hard

1.70

hard

1.63

HARD

1.50

 HARD

1.34

wired

1.31

硬

1.23

 harder

1.19

 partying

1.13

Activations Density 0.027%