Corrections after a claim-by-claim check of the errata against the findings paper and the data.
- Card: the `baselines.variant` note now says exactly what `official` and `alt` hold, that an earlier `pre_dedup_pre_enhy_fix` snapshot stays in `grand`, and which results of the findings paper can tell the two snapshots apart (only those computed from the error spans; Table 6 already fits the deduplicated snapshot).
- Card: a new warning that the GPT-5.6 Sol baseline at xHigh reasoning is incomplete (placeholder `t2_score` in every row; mostly empty Task 1 predictions for `eng-sme` and `eng-ind`).
- Field descriptions of `baselines.variant` and `systems.tasks` updated, so those Parquet files change in their schema metadata only; the data are unchanged (validation: 0 failures).
Built from commit `5a0a7dc` of the tooling: https://github.com/wmt-conference/wmt26-mteval/pull/1. `wmt26-grand.parquet` is not touched.
Adds the v0.9 typed configs and a full dataset card, alongside the existing `wmt26-grand.parquet`, which is unchanged.
**Added:** seven typed configs, each with one split named `evaluation` (the name this repository's first card and the grand file's own metadata use):
| config | rows |
|---|---:|
| `baselines` | 10,753,092 |
| `challenge_sets` | 7 |
| `items` | 26,885 |
| `mt_systems` | 55 |
| `pairs` | 363,242 |
| `submissions` | 11,297,384 |
| `systems` | 73 |
`grand` stays a config of its own (it needs `datasets>=4.7.0`), and `pairs` becomes the default config.
**README.md** is replaced by the full card. The first card's points are kept: one row per item in `grand`, nested JSON payloads, a single evaluation split not partitioned into train/test sets, and raw annotation sessions kept outside the table. The card is rendered from the data, so every number in it is counted, not typed.
**Checks run before this upload:**
- validation: 0 failures (row counts, paper Table 2, personal d…
