INDEX

Explanations

forms of 'to be' and 'to do'

The neuron fires most strongly on common one‐ and two‐letter function words—especially pronouns (I, they, it, men) and simple auxiliary verbs (is, am, do, did).

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

-1.16

-1.13

of

-1.11

de

-1.06

future

-1.04

re

-1.03

 حتی

-1.02

 samme

-1.01

roxene

-0.99

den

-0.97

POSITIVE LOGITS

zesse

1.27

 różnych

1.25

 всегда

1.22

 эти

1.16

tekend

1.16

ünkü

1.13

olyb

1.13

kve

1.13

今年は

1.11

 bây

1.11

Activations Density 0.038%