INDEX

Explanations

relative pronouns and clauses

The neuron strongly fires on content‐bearing tokens—especially nouns and names (e.g. “Philip,” “код,” “прогу,” “momентов”)—rather than on common function 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

Where

-1.42

 bahwa

-1.30

 Bedürfn

-1.22

Disliked

-1.16

erba

-1.12

倂

-1.09

を楽しむ

-1.08

ætte

-1.08

Donativos

-1.07

我也

-1.07

POSITIVE LOGITS

 which

8.06

who

7.41

 που

5.41

 that

5.38

 которая

4.84

 الذي

4.78

 которые

4.63

 که

4.41

ที่

4.38

 التي

4.38

Activations Density 0.006%