INDEX

Explanations

button with digit 9

The neuron fires most strongly on subordinating “linking” words that kick off dependent clauses (e.g. after, when, where).

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

 व्‍य

0.82

 क्‍या

0.75

ORD

0.71

 لكن

0.71

be

0.71

OR

0.71

 অপরদিকে

0.70

𝙽

0.68

={

0.68

AOS

0.67

POSITIVE LOGITS

ри

0.82

쐐

0.78

ating

0.76

 tartal

0.75

رس

0.73

אם

0.72

ählen

0.71

উপর

0.70

ச

0.70

риб

0.70

Activations Density 0.001%