INDEX

Explanations

purpose or explicit content

The neuron consistently fires on common short function words (prepositions, conjunctions, and other high‐frequency “stop” words) across these multilingual forum posts.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

 which

-0.90

}}{\

-0.87

তথ্যসূত্র

-0.86

メニューは

-0.81

 felb

-0.81

 welchem

-0.79

])

-0.79

jarige

-0.78

)}{\

-0.77

 этому

-0.76

POSITIVE LOGITS

 того

4.22

 اینکه

2.67

 том

2.47

 quello

2.45

 тому

2.34

 that

2.28

 ciò

2.20

That

2.16

 faptul

2.08

того

2.06

Activations Density 0.015%