INDEX

Explanations

introduces a hypothetical scenario

The neuron fires on question- or instruction-introducing words in math problems (e.g. “Let,” “Suppose,” “Assuming,” “What,” “Which,” “Calculate,” “Determine”).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

-1.28

 diferentes

-1.27

での

-1.22

().

-1.19

।

-1.17

abilité

-1.16

 Bewegungen

-1.14

एम

-1.13

 những

-1.12

necedor

-1.11

POSITIVE LOGITS

 this

1.61

 then

1.38

 Therefore

1.36

 This

1.35

 також

1.34

Suivez

1.31

 However

1.30

 auparavant

1.30

Télécharger

1.26

 вона

1.23

Activations Density 0.023%