INDEX

Explanations

KISS principle

The neuron strongly activates on domain-specific or proper names and technical terms (especially multi-token expressions and possessives).

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

 destin

-0.71

 autentic

-0.69

 אמ

-0.69

 tekst

-0.68

 Batch

-0.68

 магази

-0.66

diel

-0.66

 maggio

-0.66

 tela

-0.66

ameran

-0.66

POSITIVE LOGITS

Oc

2.94

Oc

2.45

 KISS

1.76

 simplest

1.60

KISS

1.49

 simpler

1.44

 simplicity

1.38

oc

1.30

 Simplicity

1.22

Simplicity

1.17

Activations Density 0.027%