INDEX

Explanations

measuring observer performance

This neuron fires on the “<” tokens (and surrounding bits like Optional or of) used in generic‐type parameter syntax. In other words, it detects generic type declarations (the opening angle‐bracket “<”).

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

-1.88

-1.81

Of

-1.69

 berühmten

-1.63

是非常

-1.51

 LOTS

-1.51

ers

-1.48

máquina

-1.46

了过来

-1.45

是很

-1.43

POSITIVE LOGITS

 what

1.95

簟

1.70

 appetite

1.69

雳

1.66

 AFFAIRS

1.65

⏃

1.61

НОЙ

1.56

eech

1.56

⃙

1.56

точный

1.52

Activations Density 0.002%