INDEX

Explanations

noun phrase pairings

The neuron fires most strongly on abstract process or technical nouns—particularly verbal noun forms (e.g. gerunds like “practice,” “staging,” “screening” or ‑tion/-ment words like “protest,” “defense,” “intervention”).

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

المناصب

-0.96

żdy

-0.93

Crest

-0.90

 kinds

-0.90

种类

-0.90

 déri

-0.89

 développ

-0.89

どこか

-0.88

 considérée

-0.87

ઃ

-0.86

POSITIVE LOGITS

 tool

2.05

for

1.83

 source

1.71

 option

1.66

 solution

1.38

 opportunity

1.37

 weapon

1.29

 contributor

1.27

aid

1.27

 vehicle

1.26

Activations Density 0.180%