INDEX

Explanations

economic feasibility

The neuron fires strongly on infrequent or “rare” tokens—e.g. unusual words, numbers, proper names, or punctuation marks—essentially detecting low‐frequency tokens.

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

shah

0.87

shortest

0.84

certificate

0.83

ValidationError

0.83

swith

0.83

cursor

0.82

Mât

0.82

 cabang

0.81

tumor

0.80

POSITIVE LOGITS

 Feelings

0.81

 방향

0.76

 NAME

0.71

 Diverse

0.70

 एच

0.69

 Doesn

0.69

 Audible

0.68

ekk

0.67

 rhyth

0.66

ੇ

0.66

Activations Density 0.000%