INDEX

Explanations

handler context

This neuron activates on occurrences of the token “handler” (and its variants) in code.

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

’

-3.27

</u>

-2.75

-2.72

</strong>

-2.47

 weißen

-2.39

苤

-2.23

”

-2.20

dijual

-2.20

acterísticas

-2.17

ísticos

-2.06

POSITIVE LOGITS

柃

3.31

孓

3.03

))

2.86

</td>

2.78

怎么说

2.73

洐

2.67

辔

2.67

镚

2.61

);

2.58

 手作り

2.58

Activations Density 0.005%