INDEX

Explanations

`=` or `and` operator

The neuron responds most strongly to standalone topic‐label or section‐heading words that mark a major content shift or new section.

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

 neigh

0.89

 تھا

0.85

ップ

0.83

岖

0.82

房价

0.82

 sagittis

0.81

 تھے

0.79

ínez

0.79

 descuentos

0.79

 ہیں

0.78

POSITIVE LOGITS

ام

0.80

Ду

0.77

config

0.75

푹

0.72

ма

0.72

ход

0.69

<%

0.69

طرف

0.69

dup

0.68

src

0.67

Activations Density 0.001%