INDEX

Explanations

opinion is

The neuron detects boilerplate or structural document headings and metadata cues (e.g. COPY, TRANSCRIPT, REVIEW, OPINION, disclaimer‐style words).

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

 that

-2.16

koliv

-1.62

גות

-1.43

"},

-1.41

----------

-1.36

ئت

-1.33

issez

-1.32

middag

-1.31

 säga

-1.27

 querían

-1.27

POSITIVE LOGITS

There

1.47

Another

1.43

1.42

 matches

1.38

 tiefer

1.34

 haven

1.33

 nieco

1.32

This

1.29

そこは

1.27

Activations Density 0.171%