EMNLP 2026

ELEVENTH CONFERENCE ON
MACHINE TRANSLATION (WMT26)

November, 2026
Budapest, Hungary
HOME

TRANSLATION TASKS: GENERAL MT •︎ INDIC MT •︎ ARABIC-ASIAN MT •︎ TERMINOLOGY •︎ MODEL COMPRESSION •︎ VIDEO SUBTITLE TRANSLATION
EVALUATION TASKS: MT TEST SUITES •︎︎ AUTOMATED MT EVALUATION
OTHER TASKS: OPEN DATA •︎ MULTILINGUAL INSTRUCTION •︎ LIMITED RESOURCES LLM

ANNOUNCEMENTS

  • 2026-04-20: Video Subtitle Translation task announced at WMT 2026

DESCRIPTION

The Video Subtitle Translation task at WMT 2026 focuses on evaluating the quality of machine translation systems in the context of video subtitle translation. Unlike general-purpose machine translation, subtitle translation presents unique challenges:

  • Length and timing constraints: Translations must be concise enough to fit on screen while matching the timing of the original subtitles.

  • Metadata utilization: Systems can leverage rich metadata including video descriptions, character lists, domain information, and glossaries.

  • Style and register: Subtitles often contain colloquial language, slang, humor, and culturally specific expressions that require careful adaptation.

The goal of this shared task is to:

  • Investigate the applicability of current MT techniques to video subtitle translation

  • Evaluate how well systems can utilize audiovisual context and metadata to improve translation quality

  • Benchmark subtitle translation quality across multiple language pairs involving Chinese

  • Encourage both academic and industry participation in advancing subtitle translation technology

We hope that both beginners and established research groups will participate in this task.

Language Pairs

The following language pairs will be evaluated (Chinese Simplified as source):

  • Chinese Simplified → English

  • Chinese Simplified → Thai

  • Chinese Simplified → Indonesian

  • Chinese Simplified → Malay

  • Chinese Simplified → Chinese Traditional (Taiwan)

IMPORTANT DATES

All dates are at the end of the day for Anywhere on Earth (AoE)

Task details finalized

20th April 2026

Test data released

1st June 2026

Translation submission deadline

1st July 2026

System description paper submission & constrained weights release

TBA August 2026

Camera-ready submission

TBA September 2026

TASK DESCRIPTION

The task is to translate video subtitles from the source language to the target language. Participants will receive subtitle source text along with rich metadata, and are expected to produce high-quality subtitle translations.

Test Data

The test data will consist of recent video subtitles with high-quality human reference translations. Each test sample will include:

  • Subtitle source text: The original subtitle text to be translated

  • Metadata: Additional context including:

    • Video description / synopsis

    • Character list (speaker names and roles)

    • Domain information (e.g., drama, documentary, news, entertainment)

    • Glossary of domain-specific terms

Participants are encouraged to make use of all available metadata to produce the best possible translations. The use of metadata is optional but may significantly improve translation quality, especially for domain-specific terminology and character names. You can also find the videos in wetv.vip and v.qq.com for reference.

Constrained Track

The Video Subtitle Translation task adopts a constrained track only, similar to the constrained track of the General MT task:

  • You are allowed to use any training data, models, and frameworks, as long as you release your model weights under a license that at least allows unrestricted non-commercial use (e.g., Apache, MIT, etc.)

  • The final model’s total number of parameters must be smaller than 30B parameters. Intermediate steps may use larger models (e.g., distillation). Mixture-of-experts counts as total parameter count. Ensembles count as the sum of their weights.

  • You are required to release the model weights together with the system submission

We recommend using Hunyuan-MT (HY-MT) as a strong base model for fine-tuning. Participants who use Hunyuan-MT as the base model and achieve competitive results may be eligible for monetary prizes. Prize details will be announced at a later date.

EVALUATION

All primary submissions will be evaluated with both automatic metrics and human evaluation. The detailed evaluation protocol will be announced together with the test data release.

Submission

Submission details will be announced together with the test data release.

DATA

Licensing. The data released for the Video Subtitle Translation task can be freely used for research purposes. We ask that you cite the shared task overview findings paper and respect any additional citation requirements on the individual data sets. For other uses of the data, you should consult with the original data owners.

CONTACT

For queries, please contact the task organizers at hunyuan@tencent.com or use the WMT mailing list.

Organizers

  • Tencent Hunyuan — hunyuan@tencent.com

  • Tencent Video