INDEX

Explanations

quantities and comparisons

The neuron flags words that express quantity or degree—i.e. magnitudes, amounts, or intensifying terms (lot/much/long/enough/etc.).

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

textit

-1.38

%).

-1.31

 pembe

-1.29

 mittler

-1.28

THA

-1.22

 distinguer

-1.20

 entspre

-1.20

mockMvc

-1.18

ossa

-1.18

 strikingly

-1.18

POSITIVE LOGITS

for

3.00

to

2.47

 that

1.78

 first

1.32

 būtų

1.29

正義

1.28

any

1.25

 případě

1.25

 begin

1.19

 Autom

1.19

Activations Density 0.017%