INDEX

Explanations

welcome or unwanted events

This neuron detects the word “welcome” (and its variants like “welcomed”) signaling a positive reception or approval.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

淠

-1.73

️⃣

-1.67

nachron

-1.58

 pié

-1.55

</strong>

-1.55

ISHED

-1.55

 erwähnten

-1.53

 virtuel

-1.52

 bahkan

-1.51

 nawet

-1.49

POSITIVE LOGITS

↵↵

2.38

1.80

and

1.75

in

1.55

却又

1.53

not

1.49

の方法

1.42

hi

1.39

ماس

1.39

et

1.38

Activations Density 0.008%