INDEX

Explanations

easy to take off

The neuron fires strongly on mentions of ease or simplicity—especially the word “easy” (often in “easy to …”) in instructional or descriptive contexts.

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

為に

-0.87

ものである

-0.86

Nef

-0.79

jshint

-0.78

kates

-0.77

BOG

-0.75

ṿ

-0.72

PDOException

-0.71

phenol

-0.71

titles

-0.71

POSITIVE LOGITS

off

0.93

off

0.80

不会

0.79

オン

0.77

put

0.73

gliere

0.72

layered

0.71

adering

0.70

 Burr

0.70

 trova

0.69

Activations Density 0.020%