INDEX

Explanations

The neuron strongly activates on sentence-initial words—especially prepositions and verbs like “In,” “On,” “At,” “For,” “working,” and “maintaining”—that introduce or transition into procedural instruction steps.

prepositions before noun phrases

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

 tênis

-1.03

事態

-0.99

 سایر

-0.96

 тому

-0.96

扩大

-0.95

齟

-0.94

 poros

-0.93

Ainsi

-0.93

 bunda

-0.92

 tirs

-0.92

POSITIVE LOGITS

 your

1.55

大概

0.98

ouchers

0.96

(''),

0.96

を始め

0.91

 tevens

0.91

idf

0.90

ępnie

0.89

 примерно

0.89

нибудь

0.89

Activations Density 0.061%