INDEX

Explanations

don't, does not, do not

This neuron fires on uppercase technical identifiers and acronyms (e.g. chemical abbreviations like QNTX or registry key names).

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

 shouldBe

-1.30

 داشته

-1.14

しかった

-1.05

頑

-1.03

[]=$

-0.96

-0.95

 HAVE

-0.94

SPECIFICATIONS

-0.94

 بوده

-0.92

非

-0.91

POSITIVE LOGITS

 doesn

5.16

don

4.84

 does

4.34

do

3.55

 δεν

3.33

 doesnt

3.23

 không

3.06

 didn

3.02

 dont

2.98

 Doesn

2.86

Activations Density 0.693%