INDEX

Explanations

array reshaping and formatting

The neuron specifically fires on file‐system and code‐related tokens—e.g. directory separators (“/”), filenames (like “models.py”, “views.py”), extensions, and related punctuation—highlighting bits of code or file paths.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

殻

0.56

ونز

0.54

 oregano

0.51

jenih

0.50

जीर

0.49

 alami

0.49

daad

0.49

 concurrently

0.49

 gleichzeitig

0.48

cec

0.48

POSITIVE LOGITS

views

0.92

 view

0.91

 views

0.89

view

0.89

urlpatterns

0.88

View

0.87

 View

0.86

VIEW

0.85

setView

0.80

 VIEW

0.80

Activations Density 0.004%