INDEX

Explanations

lock and locke variations

The neuron fires strongly on any token containing the character sequence “lock” (in any casing or as part of a larger word).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

.$.

-0.79

 effets

-0.76

ĝ

-0.76

唱歌

-0.73

decks

-0.69

újo

-0.69

ಡ

-0.67

 shifts

-0.67

 brancas

-0.67

 mágica

-0.66

POSITIVE LOGITS

heed

1.09

yer

1.02

 Lock

0.94

Lok

0.91

 LOCK

0.81

 lock

0.81

föra

0.81

Lock

0.80

 Gustaf

0.80

Loire

0.79

Activations Density 0.014%