INDEX

Explanations

superlative descriptions

This neuron detects superlative or high‐degree evaluative adjectives (and related ranking words like “best,” “most,” “greatest,” “known,” etc.).

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

of

-2.39

you

-1.61

 they

-1.33

 nuances

-1.19

も含

-1.16

 remainder

-1.08

zegovina

-1.06

 gleiche

-1.05

 amount

-1.05

洇

-1.04

POSITIVE LOGITS

own

1.60

最も

1.38

Caratteristiche

1.35

كري

1.27

Both

1.26

بري

1.24

Specifiche

1.23

！");

1.23

Źródło

1.22

ământ

1.19

Activations Density 0.023%