INDEX

Explanations

The neuron lights up on the main content words in explanatory or descriptive passages—i.e. the key nouns and verbs that carry factual information rather than on function words or dialog.

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

それでも

0.57

્યે

0.56

 concerted

0.55

aim

0.54

 اقدامات

0.53

 fully

0.52

 underwent

0.52

跪

0.52

 exhaustive

0.51

ดำเนิน

0.51

POSITIVE LOGITS

 Improves

1.41

 improves

1.25

 улуч

1.23

 enhances

1.21

 Reduces

1.20

 strengthens

1.20

 relieves

1.19

 améliorer

1.17

 meningkatkan

1.15

减轻

1.13

Activations Density 0.897%