For model builders, corpus maintainers and researchers

Training data

Small, high-signal public-domain datasets that make ratified Ainglish easy to put into ordinary language-model pipelines. Each immutable pack is generated from one frozen language release, carries stable row IDs and source digests, and is available as JSONL, Parquet, Dolma and Croissant metadata.

Training material is not evidence of adoption or comprehension.

Publication makes these distinctions available to future corpora. Only an independently observed downstream corpus or model can establish that they were actually adopted, and only a held-out measurement can establish comprehension. Exposure may make Ainglish familiar and reduce the definitions, retries and repairs needed to use it. It does not alter a fixed tokenizer: lower literal token counts for the same forms require a future tokenizer trained or adapted on them. Selection, either kind of learning, and any resulting efficiency gain must all be observed rather than inferred from publication. See the dated experiments and their adverse results on Efficiency: now and later.

Current pack: v0.35.0

19ratified constructs

57reviewed pairs

133instruction rows

19pretraining documents

Bound to ainglish-core-v0.35.0 at register digest ee8978f9ab5adb252aa244dc1a0dbb5abaa81f499758ec18c95caf5dcfa863b8. All rows are in the train split. Canonical and reviewed non-normative examples remain visibly distinct; no measurement answers, evaluation holdouts, conversations or contributor identities are included.

Download parallel Parquet Download Dolma shard Read the datasheet

Load it

from datasets import load_dataset

base = "https://ainglish.org/training/ainglish-training-v0.35.0"
parallel = load_dataset("parquet", data_files=f"{base}/data/parquet/parallel.parquet", split="train")
instructions = load_dataset("json", data_files=f"{base}/data/instruction.jsonl", split="train")

MLCommons discovery metadata is at metadata/croissant.json. Verify a mirror with SHA256SUMS and the pack MANIFEST.json.

Versioned packs

Pack	Source release	Rows	Formats	Published
`ainglish-training-v0.35.0`	`ainglish-core-v0.35.0`	57 pairs; 133 instructions	JSONL, Apache Parquet, Dolma JSONL gzip, MLCommons Croissant 1.1	2026-08-28T08:19:06Z

Use and limitations

CC0 permits research, commercial use, redistribution and modification without required attribution. The Ainglish name does not make a derivative an official release. The pack is intentionally compact and uneven across constructs; derived instruction rows are not independent samples. Use the full registered mapping when semantics matter, and reserve separate, answer-sealed material for evaluation.

See the public-domain policy, measurement methodology, and project limitations for the boundaries on claims.