INDEX

Explanations

This neuron is effectively inactive (a “dead” neuron) and does not detect any pattern in the input.

The neuron detects discourse-structuring / transition tokens—words and short phrases used to introduce sections, disclaimers, or meta-commentary (e.g., "Okay", "Before", "Instead", "Adding").

New Auto-Interp

Configuration

Prompts (Dashboard)

238,145 prompts, 512 tokens each

Dataset (Dashboard)

lmsys + oasst1

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

ckså

0.27

ewöhn

0.26

myCollision

0.26

 ਉਹ

0.25

ukkig

0.25

okolade

0.25

Ᏸ

0.25

 সম্ভবত

0.25

ంచెం

0.25

beatCounter

0.25

POSITIVE LOGITS

 according

0.28

 beserta

0.27

0.26

 provides

0.26

0.24

0.23

 countries

0.23

 includes

0.23

Activations Density 0.000%