INDEX

Explanations

contrastive conjunctions and adverbs

This neuron detects hyphenated or split word fragments (subword pieces joined or broken by “-”).

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

common

0.54

($"{

0.54

resizable

0.51

fillna

0.49

consider

0.48

substr

0.48

readonly

0.47

сту

0.47

pred

0.47

দেখ

0.46

POSITIVE LOGITS

But

0.65

 They

0.59

 Adresse

0.59

 AGRICULTURE

0.59

 แต่

0.59

 परन्तु

0.59

érist

0.58

 lessee

0.58

And

0.57

 Nhưng

0.56

Activations Density 0.004%