INDEX

Explanations

leaving or staying

This neuron detects formal discourse markers—specifically words like “shall” (a formal future/modal marker) and “two” (an enumeration marker) in formal, doctrinal text.

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

𝐴

0.83

abras

0.82

ών

0.80

𝐺

0.79

Consent

0.77

𝑚

0.77

 змін

0.76



0.75

 sintomi

0.74

abuse

0.74

POSITIVE LOGITS

しっかり

0.86

し

0.83

Bl

0.73

ل

0.72

 seperti

0.71

 padx

0.71

panda

0.70

 puffed

0.70

/>}

0.69

ו

0.67

Activations Density 0.000%