Benchmark Almanac

UD Latin PROIEL

HIT-SCIR (Harbin) 73.6% tops its leaderboard
Not in model cards

Models analyze Latin sentences from the Vulgate and from authors including Caesar and Cicero. The test includes both biblical and secular prose.

Bamman and Burns tested Latin BERT on part of speech tagging. Kupari and colleagues also evaluated grammatical dependencies on this treebank.

Released
November 2015
Kind
Reading in many languages
Level
University
Scored
Checked automatically
Answers
Short answer
Tasks
All public
Measures
Capability
Language
Latin
In and out
Text → Text
Interaction
Single turn
Items from
Real work and use
Run by
Anyone
26
Entries on its own leaderboard, 0 models we track

HIT-SCIR (Harbin) tops the leaderboard with 73.6%

Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders.

  1. 1 HIT-SCIR (Harbin) software1-P 73.6%
  2. 2 Stanford (Stanford) software2 71.8%
  3. 3 TurkuNLP (Turku) software1-P 71.5%
  4. 4 CEA LIST (Paris) software1-P 71.1%
  5. 5 ICS PAS (Warszawa) software1-P 70.1%
  6. 6 UDPipe Future (Praha) software1-P 69.8%
  7. 7 ParisNLP (Paris) software1-P 67.9%
  8. 8 LATTICE (Paris) software1-P 67.7%
  9. 9 NLP-Cube (București) software1-P 67.4%
  10. 10 SLT-Interactions (Bengaluru) software2-P 67%
  11. 11 iParse (Pittsburgh) software1-P 66.4%
  12. 12 AntNLP (Shanghai) software1-P 66.2%
  13. 13 IBM NY (Yorktown Heights) software1-P 64.7%
  14. 14 Uppsala (Uppsala) software1-P 64.1%
  15. 15 Fudan (Shanghai) software5-P 62.6%
  16. 16 LeisureX (Shanghai) software1-P 62.6%
  17. 17 UniMelb (Melbourne) software1-P 62%
  18. 18 Phoenix (Shanghai) software1-P 60.4%
  19. 19 BASELINE UDPipe 1.2 (Praha) software1-P 59.7%
  20. 20 KParse (İstanbul) software1-P 59.4%
  21. 21 BOUN (İstanbul) software2-P 58.2%
  22. 22 CUNI x-ling (Praha) software1-P 55.2%
  23. 23 HUJI (Yerushalayim) software1-P 53%
  24. 24 ArmParser (Yerevan) software1-P 52%
  25. 25 ONLP lab (Ra'anana) software3-P 48.1%
  26. 26 SParse (İstanbul) software1-P 0%

Source: Official UD Latin PROIEL leaderboard

#EntryScoreSettingDateSource
1 HIT-SCIR (Harbin) 73.6% software1-P Official leaderboard
2 Stanford (Stanford) 71.8% software2 Official leaderboard
3 TurkuNLP (Turku) 71.5% software1-P Official leaderboard
4 CEA LIST (Paris) 71.1% software1-P Official leaderboard
5 ICS PAS (Warszawa) 70.1% software1-P Official leaderboard
6 UDPipe Future (Praha) 69.8% software1-P Official leaderboard
7 ParisNLP (Paris) 67.9% software1-P Official leaderboard
8 LATTICE (Paris) 67.7% software1-P Official leaderboard
9 NLP-Cube (București) 67.4% software1-P Official leaderboard
10 SLT-Interactions (Bengaluru) 67% software2-P Official leaderboard
11 iParse (Pittsburgh) 66.4% software1-P Official leaderboard
12 AntNLP (Shanghai) 66.2% software1-P Official leaderboard
13 IBM NY (Yorktown Heights) 64.7% software1-P Official leaderboard
14 Uppsala (Uppsala) 64.1% software1-P Official leaderboard
15 Fudan (Shanghai) 62.6% software5-P Official leaderboard
16 LeisureX (Shanghai) 62.6% software1-P Official leaderboard
17 UniMelb (Melbourne) 62% software1-P Official leaderboard
18 Phoenix (Shanghai) 60.4% software1-P Official leaderboard
19 BASELINE UDPipe 1.2 (Praha) 59.7% software1-P Official leaderboard
20 KParse (İstanbul) 59.4% software1-P Official leaderboard
21 BOUN (İstanbul) 58.2% software2-P Official leaderboard
22 CUNI x-ling (Praha) 55.2% software1-P Official leaderboard
23 HUJI (Yerushalayim) 53% software1-P Official leaderboard
24 ArmParser (Yerevan) 52% software1-P Official leaderboard
25 ONLP lab (Ra'anana) 48.1% software3-P Official leaderboard
26 SParse (İstanbul) 0% software1-P Official leaderboard

No lab model card reports it yet

We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.

More in Multilingual

All 545 in Multilingual
  • MMMLU In use MMLU professionally translated into 14 languages. Multiple-choice accuracy. Claude Mythos Preview 92.7%
  • WMT24++ In use English-to-55-language translation of WMT24 test data. Scored with automatic metrics such as MetricX or xCOMET. K-EXAONE 90.5%
  • MMLU-ProX In use MMLU-Pro translated into up to 29 languages with expert review. Multiple-choice accuracy per language. Qwen3.7-Max 87%
  • CLIcK In use Models answer 1,995 multiple-choice questions on Korean language and culture. Questions come from official Korean exams and textbooks in eleven categories. Solar Open 2 90.7%
  • Global MMLU Lite 3.0 In use Models answer 400 human-translated MMLU questions per language in 23 languages. Version 3.0 adds Czech, Hungarian, Oriya, Slovak and Tajik and updates Italian. Accuracy. Claude Opus 5 95.7%
  • MMMB In use Models answer multiple-choice questions about images in six languages across 15 categories. The set has about 12,000 questions. LFM2.5-VL-3B 83%

Data audit

All 8 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 2 pass 3 notes
  • Pass Independent runs Scores from the official leaderboard (26 entries).
  • Pass Models scored 26 entries on its leaderboard, none a model we track.
  • Note Newest score No scores yet.
  • Note Frontier models No models scored yet.
  • Note Headroom No scores yet.
Leaderboard import 3 pass
  • Pass Importer We read the board on universaldependencies.org: 26 entries at the last read, none a model we track.
  • Pass Last read November 20th, 2026.
  • Pass Read worked The last read worked.
Consistency 1 pass 1 note
  • Note Lab claims No model has both a lab number and an independent run.
  • Pass Known issues No published audit found problems.

How we check