INDEX

Explanations

time indicators: HH:MM, AM, per night

The neuron strongly activates on rare or unexpected tokens—especially proper names, specialized terms, and other low‐frequency 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

ᆹ

0.88

ляць

0.84

kipun

0.82



0.81

 Oaks

0.80

}_{+}^{

0.80

├──

0.79

 કહ્યું

0.78

 METHODS

0.78

 університе

0.77

POSITIVE LOGITS

 Elle

0.73

↵

0.71

기는

0.71

ally

0.70

anged

0.69

Є

0.68

elle

0.67

 Angebot

0.66

т

0.66

 what

0.65

Activations Density 0.000%