INDEX

Explanations

prepositions followed by determiners/pronouns

The neuron selectively fires on technical/programming discourse, especially on verbs and terms that describe coding activities (e.g. “writing,” “regarding,” “adhere,” “understand”) in Q&A–style posts.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 quien

-1.41

who

-1.33



-1.14

 pernic

-1.12

who

-1.12



-1.10

 Turbul

-1.09

 diagon

-1.08

ftet

-1.08

how

-1.07

POSITIVE LOGITS

 this

1.61

 their

1.32

his

1.25

its

1.22

ruguay

1.13

1.10

 tillgäng

1.09

estä

1.05

 argentinos

1.05

uuuu

1.04

Activations Density 0.087%