INDEX

Explanations

slur with ring or Institutes with of

This neuron detects when a meaningful word stem is split across two tokens—e.g. “slur” + “ring,” “man” + “hole,” or “slur” + “ries”—marking morpheme boundaries in compound or inflected words.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

-2.63

и

-2.20

they

-2.11

𝐲

-2.11

和

-1.98

鈁

-1.98

その

-1.95

⚣

-1.93

 parachoque

-1.90

᠁

-1.88

POSITIVE LOGITS

er

2.44

鲵

2.34

いえば

2.19

ar

2.14

して下さい

2.06

言えば

1.99

 nears

1.95

컫

1.94

AISSEE

1.92

姐妹

1.91

Activations Density 0.015%