INDEX

Explanations

source in code

This neuron specifically detects occurrences of the token “source.”

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

who

-1.52

 whom

-1.52

umacher

-1.51

and

-1.38

 покупать

-1.34

-1.29

 credere

-1.29

-1.28

},

-1.27

 sabias

-1.27

POSITIVE LOGITS

 ۱۷

1.63

ниці

1.58

𐐪

1.52

валі

1.51

плі

1.51

 sebuah

1.50

当事

1.49

 ένα

1.48

蓑

1.47

mètre

1.46

Activations Density 0.061%