INDEX

Explanations

This neuron responds to instructional or explanatory language, flagging words that signal explanations, reasoning, or guidance (e.g. “explanation,” “reasoning,” “recommend,” “move through”).

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 Shivaji

0.58

 notícia

0.56

েম্ব

0.55

 müm

0.55

 खबरें

0.54

 revêt

0.54

 acudir

0.54

 notícias

0.54

 Сурикова

0.53

 Standort

0.52

POSITIVE LOGITS

 reasoning

1.92

 explanation

1.76

explanation

1.72

 explanations

1.65

 justifications

1.61

Explanation

1.60

 justification

1.60

 rationale

1.59

 Reasoning

1.59

 Explanation

1.56

Activations Density 0.169%