INDEX

Explanations

liking, wanting, or expecting

The neuron flags words that express disagreement or negative sentiment (e.g. “displeases,” “or,” “like” in a negative context).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

to

-1.52

 ролях

-1.37

After

-1.25

now

-1.23

-1.21

Regardless

-1.20

sellors

-1.20

During

-1.19

</sup>

-1.18

 даже

-1.17

POSITIVE LOGITS

の結果

1.45

 where

1.43

 trzec

1.36

 trovare

1.33

 choices

1.33

 what

1.32

muñeca

1.30

拊

1.25

 חלק

1.23

 afir

1.22

Activations Density 0.050%