INDEX

Explanations

hypothesis, problem, project, affair

The neuron fires on technical keywords that denote major conceptual elements in scientific paper titles or headings—words like “effect,” “hypothesis,” “problem,” “phenomenon,” and “paradox.”

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

of

-1.88

 with

-1.45

 įva

-1.43

 entweder

-1.43

 unimp

-1.41

Đây

-1.35

 такие

-1.32

 水着

-1.30

 such

-1.30

 этой

-1.30

POSITIVE LOGITS

叕

1.38

","+

1.36

[]):

1.36

 adulta

1.28

ᴀ

1.25

 teori

1.24

 universitaria

1.23

 gana

1.23

زال

1.21

 brisa

1.21

Activations Density 0.167%