INDEX

Explanations

rewrite the

The neuron strongly activates on capitalized “Re-” name fragments (e.g. Rebolledo, Reimann, Rebis, Regev) as they appear in bibliographic author lists.

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

öm

0.60

俨

0.57

광역시

0.55

essero

0.55

儼

0.54

ிருந்து

0.53

 الحاله

0.53

 शाय

0.52

lemish

0.52

 Gestion

0.52

POSITIVE LOGITS

Re

1.14

re

1.07

Re

1.02

RE

0.95

 reconfiguration

0.88

re

0.87

 ре

0.85

 recombin

0.82

Rew

0.81

 recomb

0.80

Activations Density 0.442%