INDEX

Explanations

prior actions or roles

The neuron strongly activates on the word “previously,” signaling prior or earlier events or roles.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

</h1>

-1.71

</strong>

-1.61

ناول

-1.58

 With

-1.55

 Both

-1.55

 when

-1.52

 This

-1.52

 Terkait

-1.49

 inicia

-1.48

誅

-1.45

POSITIVE LOGITS

谂

1.93

ᜀ

1.90

鿬

1.89

 anhänger

1.87

previously

1.84



1.77

ts

1.77

わたしは

1.75

䛗

1.74

莫名

1.73

Activations Density 0.026%