INDEX

Explanations

relative clauses

This neuron detects the relative pronoun “which” (and related linking/function words introducing clauses).

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

 напряжение

-1.50

tifications

-1.48

-1.46

Имя

-1.45

yor

-1.45

熒

-1.40

 Bisa

-1.38

烤

-1.38

),

-1.37

ᆞ

-1.34

POSITIVE LOGITS

The

1.89

 nuevas

1.77

禤

1.59

稣

1.58

ívne

1.55

 sonre

1.53

 fueron

1.51

 najwięks

1.48

 porém

1.48

soo

1.48

Activations Density 0.036%