INDEX

Explanations

options and desires

This neuron lights up on semantically “heavy” content words (not common function words), i.e. tokens that carry real topical or descriptive weight.

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

 почти

0.77

 фактически

0.69

 майже

0.65

Almost

0.63

 گئے۔

0.61

 prawie

0.60

 계속

0.59

 quase

0.58

 debacle

0.58

 கிட்டத்தட்ட

0.58

POSITIVE LOGITS

 хотите

1.27

 ingin

1.23

 muốn

1.20

 hoặc

1.13

 અથવા

1.10

または

1.09

 aspiring

1.09

 chcete

1.08

 interested

1.07

 atau

1.05

Activations Density 1.521%