INDEX

Explanations

good vs. evil conflict

The neuron flags structural “info‐header” tokens—dates, numerals, and question‐word cues (how/what/where/when/why) that commonly introduce factual or instructional content.

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

 '*',

0.57

 වන්නේ

0.57

Favorites

0.56

:*

0.56

städ

0.55

ahar

0.55

{},

0.54

Logout

0.54

ίζεται

0.54

 እነዚህ

0.54

POSITIVE LOGITS

 শীর্ষক

1.22

 అనే

1.11

という

1.09

 kind

1.04

 syndrome

1.04

 kinda

1.02

 című

1.01

 라는

0.96

 adlı

0.95

 sort

0.95

Activations Density 0.588%