INDEX

Explanations

shows

The neuron fires on the rhetorical “Here/We …” and “In this study/We …” style phrases that introduce a paper’s key methods or findings in scientific abstracts and articles.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 iteratively

0.93

 variational

0.87

 methodological

0.86

 researches

0.86

 combinatorial

0.84

 uniaxial

0.84

 MATLAB

0.82

 applicability

0.82

 anisotropic

0.82

 ഗവേഷ

0.82

POSITIVE LOGITS

 assh

0.76

 estómago

0.75

 milhões

0.74

 millones

0.73

道歉

0.73

 asesin

0.72

 escánd

0.72

 showbiz

0.71

 dinheiro

0.71

ocalypse

0.71

Activations Density 0.061%