INDEX

Explanations

movement and direction

The neuron detects phrasal‐verb constructions—verbs paired with a following particle (e.g. “back,” “let go,” “hold on,” “reach out”).

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

鹀

-2.19

 другими

-1.95

 decorativo

-1.84

 éché

-1.82

 múltiple

-1.75

 rafraî

-1.75

🈶

-1.75

 réfrig

-1.73

 jeśli

-1.70

are

-1.68

POSITIVE LOGITS

for

2.67

of

2.52

 also

2.45

in

2.33

 about

2.17

 unſ

1.91

 arī

1.90

and

1.88

to

1.86

 from

1.80

Activations Density 0.039%