INDEX

Explanations

model code syntax

The neuron activates on numeric literals (especially decimal numbers) in the 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

This

-1.61

<h2>

-1.53

<b>

-1.49

 erstmal

-1.45

/**

-1.38

There

-1.37



-1.35

 '+':

-1.32

')[

-1.31

。

-1.28

POSITIVE LOGITS

 akik

1.45

 kteří

1.43

makeat

1.41

揶

1.37

 seksi

1.36

hogy

1.30

∖

1.27

 saati

1.27

offering

1.27

whom

1.27

Activations Density 0.001%