INDEX

Explanations

testing assertions

This neuron activates on numeric sequences in code or comments, i.e. literal numbers.

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

 tests

-0.89

」,

-0.84

zehn

-0.83

ziem

-0.82

 test

-0.82

 their

-0.79

acceptable

-0.79

ugier

-0.77

fehl

-0.76

 romántico

-0.76

POSITIVE LOGITS

 believe

1.23

assertEquals

1.12

 Believe

0.94

mark

0.91

 belief

0.90

Believe

0.87

assertNotNull

0.86

 beliefs

0.84

assert

0.81

atterns

0.78

Activations Density 0.001%