INDEX

Explanations

the verb 'to see'

The neuron strongly activates on occurrences of the verb “see” (especially its past‐tense and past‐participle forms like “saw” and “seen”).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

︲

-2.09

 diminu

-1.66

があれば

-1.65

をつけて

-1.62

several

-1.60

を変える

-1.59

一堆

-1.56

having

-1.56

 DEPOS

-1.56

螃

-1.55

POSITIVE LOGITS

 When

1.84

 когда

1.77

</strong>

1.77

 This

1.70

</h3>

1.64

↵↵

1.53

 или

1.51

 Ironically

1.47

When

1.46

Activations Density 0.015%