INDEX

Explanations

substitution:

The neuron strongly activates on weighty abstract nouns denoting formal social, legal, or medical processes or events (e.g. resettlement, detention, rehabilitation, abduction, suicide, abortion).

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

ड्स

0.56

Tik

0.54

 Multif

0.53

Things

0.52

 THINGS

0.52

formas

0.52

чей

0.52

甶

0.51

Zus

0.51

 Plumbing

0.51

POSITIVE LOGITS

 clause

0.73

 syndrome

0.66

主义

0.61

 синд

0.61

clause

0.60

वादियों

0.60

霜

0.57

ism

0.57

 disillusion

0.56

 Clause

0.56

Activations Density 0.393%