INDEX

Explanations

context only

This neuron detects section‐header tokens that introduce or label parts of a prompt (e.g. “CONTEXT,” “TASK,” “Extract,” “Read,” “text”).

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

 bayi

0.43

 progetto

0.42

Selections

0.42

 deseos

0.41

 superiores

0.41

 मुकेश

0.41

 projeto

0.40

 proyecto

0.40

ركه

0.40

 töö

0.40

POSITIVE LOGITS

 Login

0.39

 Hotels

0.37

 जिल्

0.36

 झलक

0.36

enable

0.35

மது

0.35

 Accountability

0.35

isInt

0.34

⛴

0.34

 Jurisdiction

0.34

Activations Density 0.014%