INDEX

Explanations

Core tip

The neuron fires on hexadecimal‐digit tokens within URL percent‐encodings or Unicode escape sequences (e.g. the “4” in “%2B” or the “7” in “%7B”).

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

of

-2.28

our

-1.77

 this

-1.67

 where

-1.61

 achieving

-1.59

 acquiring

-1.55

 usual

-1.49

 that

-1.48

-1.47

 through

-1.45

POSITIVE LOGITS

<bos>

1.70



1.60

 beiden

1.59

apkan

1.58

 lmao

1.53

开始了

1.52

entrer

1.52

ichteten

1.52

eeee

1.51

 две

1.50

Activations Density 0.007%