---
license: cc-by-4.0

configs:
- config_name: wmt2020-2022
  data_files: "mqm-wmt-2020-2022.csv"
  default: true
- config_name: wmt2023
  data_files: "mqm-wmt-2023.csv"
- config_name: wmt2024
  data_files: "mqm-wmt-2024.csv"
---


This dataset contains all MQM human annotations from [WMT Metrics Shared Tasks](https://wmt-metrics-task.github.io/) from 2020 to 2024.

The data is organized into different multiple columns, all should contain the following columns:

- lp: language pair
- src: input text
- mt: translation
- ref: reference translation

This dataset is taken directly from the original [WMT repository](https://github.com/google/wmt-mqm-human-evaluation).