Benchmark Almanac

XOR-TyDi QA

PrimeQA (DrDecr-large with PLAID + Colbert V2) 74.7% tops its leaderboard
Not in model cards

Models retrieve passages across languages to answer questions written by speakers of seven languages. Tasks require English evidence or answers in the question’s language.

The authors use version 1.1 for leaderboard submissions. They report retrieval and answer quality separately.

Made by
Akari Asai and colleagues
Released
October 2020
Kind
Reading in many languages
Level
School
Scored
Checked automatically
Answers
Short answer
Tasks
All public
Measures
Capability
Language
Many languages
In and out
Text → Text
Interaction
Single turn
Run by
Anyone
Based on
TydiQA
In views
Many languages
15
Entries on its own leaderboard, 0 models we track

PrimeQA (DrDecr-large with PLAID + Colbert V2) tops the leaderboard with 74.7%

Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. The headline is XOR-Retrieve recall at 5,000 tokens, averaged over seven languages. 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 PrimeQA (DrDecr-large with PLAID + Colbert V2) White-box 74.7%
  2. 2 Quick White-box 72%
  3. 3 PrimeQA (DrDecr with PLAID + Colbert V2) White-box 71.9%
  4. 4 GAAMA (ColBERT Ensemble with IBM NMT + Google MT) External APIs 71.4%
  5. 5 LEXA White-box 70.5%
  6. 6 DrDecr White-box 70.3%
  7. 7 DPR + Google Translate External APIs 67.2%
  8. 8 Sentri 2.0 base White-box 64.6%
  9. 9 Contrastive Context-aware Pretraining Model (CCP) White-box 63%
  10. 10 Path Retriever + Google Translate External APIs 61.7%
  11. 11 Single Encoder Retriever (Sentri) White-box 61%
  12. 12 Single Encoder Retriever (Sentri, resubmission) White-box 60.7%
  13. 13 GAAMA (ColBERT ensemble with xlm-r + UW Translate) White-box 59.9%
  14. 14 DPR + Vanilla Transformer MT White-box 50%
  15. 15 Multilingual DPR White-box 48%

Source: Official XOR-TyDi QA leaderboard

#EntryScoreSettingSubmitted byDateSource
1 PrimeQA (DrDecr-large with PLAID + Colbert V2) 74.7% White-box IBM Research AI Oct 2022 Official leaderboard
2 Quick 72% White-box Microsoft STCA Jun 2022 Official leaderboard
3 PrimeQA (DrDecr with PLAID + Colbert V2) 71.9% White-box IBM Research AI Aug 2022 Official leaderboard
4 GAAMA (ColBERT Ensemble with IBM NMT + Google MT) 71.4% External APIs IBM Research AI, NY Jun 2021 Official leaderboard
5 LEXA 70.5% White-box Huawei Noah's Ark lab Jun 2022 Official leaderboard
6 DrDecr 70.3% White-box IBM Research AI Feb 2022 Official leaderboard
7 DPR + Google Translate 67.2% External APIs University of Washington, AI2, Google, UT Austin Apr 2021 Official leaderboard
8 Sentri 2.0 base 64.6% White-box Huawei Noah's Ark lab Mar 2022 Official leaderboard
9 Contrastive Context-aware Pretraining Model (CCP) 63% White-box Microsoft STCA Jan 2022 Official leaderboard
10 Path Retriever + Google Translate 61.7% External APIs University of Washington, AI2, Google, UT Austin Apr 2021 Official leaderboard
11 Single Encoder Retriever (Sentri) 61% White-box Huawei Noah's Ark lab Aug 2021 Official leaderboard
12 Single Encoder Retriever (Sentri, resubmission) 60.7% White-box Huawei Noah's Ark lab Oct 2021 Official leaderboard
13 GAAMA (ColBERT ensemble with xlm-r + UW Translate) 59.9% White-box IBM Research AI, NY Jun 2021 Official leaderboard
14 DPR + Vanilla Transformer MT 50% White-box University of Washington, AI2, Google, UT Austin Apr 2021 Official leaderboard
15 Multilingual DPR 48% White-box University of Washington, AI2, Google, UT Austin Apr 2021 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 (15 entries).
  • Pass Models scored 15 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 nlp.cs.washington.edu: 15 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