INDEX

Explanations

tokens like 'θ'

The neuron responds most strongly to low‐frequency or unusually specific tokens—that is, rare or outlier words.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

 anemia

0.88

ապ

0.80

кры

0.80

 antibiotics

0.79

 opioid

0.79

 seaweed

0.77

ダメージ

0.76

𝓈

0.76

 opioids

0.76

 swamps

0.74

POSITIVE LOGITS

 notamment

0.79

 nécessaire

0.79

sic

0.76

 emprend

0.74

 livre

0.73

 inol

0.73

 seule

0.73

élect

0.71

 rapporte

0.71

Activations Density 0.000%