INDEX

Explanations

complex or specific topics

The neuron looks for instructional or how-to style language—especially second-person cues and guiding words (e.g. “you,” “if,” “instead,” “see,” “data”) that signal procedural explanations.

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

ApJ

-0.96

llllll

-0.96

 enemigos

-0.94

haft

-0.92

bars

-0.91

ｓ

-0.91

 '/';

-0.91

lllllll

-0.91

,</

-0.91

="/">

-0.90

POSITIVE LOGITS

 complicated

1.13

 complex

1.07

but

1.04

 topics

1.02

 complicate

0.96

应该

0.93

大変

0.91

 topic

0.91

 комплекс

0.90

อีก

0.89

Activations Density 0.054%