Languages
How good is it on decisions that aren't in English?
The same kind of decision as the English edit review track, on four other Wikipedias.
Non-English accuracy German, Spanish, French, Japanese
Jev 1.14
TypeSafe
Jev 1.13
TypeSafe
Clef
Cloudflare
Clef-Flash
Cloudflare
pplx-decider v1.2
Perplexity
pplx-decider-flash v1.2
Perplexity
OpenAI Decisions
OpenAI
Microsoft-Decision-1
Microsoft
Kev-4B
Jared Palmer
Accuracy on edit review decisions from German, Spanish, French and Japanese Wikipedia, pooled.
Test items and precision
- Items
- 300 edit review decisions in each of German, Spanish, French and Japanese per test set
- Precision
- About ±5 points per language; gaps to English need about 7 points to show.