INDEX

Explanations

comparisons with extreme, greater, less

The neuron is tuned to pick out metadata field names or section‐header tokens (like “Product,” “tools,” “agent,” “Related Post,” numerical labels, etc.) rather than regular prose words.

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

ן

0.79

 частина

0.78

צב

0.75

昨年

0.73

Fuente

0.73

ר

0.72

رة

0.71

区别

0.70

ি

0.70

Ils

0.70

POSITIVE LOGITS

 foliage

0.96

 inclement

0.87

 flickering

0.86

 streamers

0.85

fand

0.84

 gimbal

0.84

 downtime

0.82

 Tidal

0.82

 lám

0.82

 canyons

0.81

Activations Density 0.001%