Blame for README.md

---
license: apache-2.0
task_categories:
  - text-generation
language:
  - en
pretty_name: Domyn Edge Teacher Math Mix
size_categories:
  - 10K<n<100K
tags:
  - domyn
  - domyn-edge
  - teacher
  - math
  - grpo
configs:
  - config_name: rl
    data_files:
      - split: train
        path: data/rl/train-*.parquet
      - split: test
        path: data/rl/test-*.parquet
---

<p align="center">
  <picture>
    <source media="(prefers-color-scheme: dark)" srcset="https://cdn.prod.website-files.com/682dcb35e7fd7a313cae803a/6835cb0fb6bd22a7cba7d645_domyn-logo-payoff-primary-white.svg">
    <img alt="Domyn" src="https://cdn.prod.website-files.com/682dcb35e7fd7a313cae803a/6835ca86622d317630fa3861_domyn-logo-primary.svg" width="400">
  </picture>
</p>

# Domyn Edge Teacher Math Mix

> [!IMPORTANT]
> **Phased release.** Domyn Edge is a rolling release in three bundles. This dataset is part of the
> second bundle, together with the domain teachers and their post-training data. For an overview,
> see the [launch blog post](https://domyn.webflow.io/blog/domyn-edge-compact-model-open).

The training data of [Domyn Edge Teacher Math](https://huggingface.co/domyn/Domyn-Edge-Teacher-Math),
in one config:

| Config | Stage | Train | Test |
|---|---|---:|---:|
| `rl` | GRPO, math problems with a verifiable final answer | 20,153 (2.96M tokens) | 815 (0.11M tokens) |

## `rl`: GRPO problems

Math problems with a single verifiable final answer. The final answer in the
response is checked against `reward.ground_truth.answer` with math-verify. Each
prompt carries a system message that asks for the final answer in `\boxed{}` or
on a final `ANSWER: $ANSWER` line.

| Source | License | Train rows | Train tokens | Test rows | Test tokens |
|---|---|---:|---:|---:|---:|
| [DeepScaleR-Preview-Dataset](https://huggingface.co/datasets/agentica-org/DeepScaleR-Preview-Dataset) | MIT | 10,508 | 1.42M | 421 | 0.05M |
| [Nemotron-3-Nano-RL-Training-Blend](https://huggingface.co/datasets/nvidia/Nemotron-3-Nano-RL-Training-Blend), math slice | ODC-BY 1.0 | 9,645 | 1.53M | 394 | 0.06M |
| **Total** | | **20,153** | **2.96M** | **815** | **0.11M** |

Tokens are prompt tokens, as in the `token_length` column. The 4,606 train rows that ask for an
`ANSWER:` line are copies of other train rows; `metadata.augmented_from` names the original.

## Decontamination

Before training, the problems were decontaminated against the benchmarks we evaluate on: AIME 2024,
AIME 2025, AIME 2026, HMMT February 2025, HMMT February 2026, GSM8K, MGSM,
GPQA Diamond, IMO-AnswerBench, IMO-ProofBench and LiveBench (math). This removed 22 train and 2 test problems. The `ANSWER:` copies were made afterwards, from
the remaining problems.

## How to use

```python
from datasets import load_dataset

rl_train = load_dataset("domyn/Domyn-Edge-Teacher-Math-Mix", "rl", split="train")
rl_test = load_dataset("domyn/Domyn-Edge-Teacher-Math-Mix", "rl", split="test")
```

## Columns

| Column | Type | Description |
|---|---|---|
| `messages` | list | The prompt: `role` and `content` of each message (the other message fields are null). |
| `tools` | string | Always null in this config. |
| `task` | string | Task family. |
| `source` | string | Upstream dataset. |
| `license` | string | License of the upstream record. |
| `metadata` | string | JSON-encoded upstream metadata. |
| `reward` | string | JSON-encoded reward data: `ground_truth` and the scoring `style`. Here `ground_truth.answer` is the reference final answer. |
| `token_length` | int | Prompt length in tokens. |
| `uuid` | string | Record identifier. |

To train with NeMo-Gym, as we did, map each row to the `math_with_judge_simple_agent` input:
`responses_create_params.input` = `messages`; `question` = `reward.ground_truth.question`, or the
user message when absent; `expected_answer` = `reward.ground_truth.answer`; `agent_ref` =
`{"type": "responses_api_agents", "name": "math_with_judge_simple_agent"}`.

## License

Our contributions to this dataset, including the curation and the mix composition, are released
under the Apache License 2.0. Upstream datasets keep their original license terms. Read
[ATTRIBUTION.md](ATTRIBUTION.md) before redistributing any part of this data.

## Citation

```bibtex
@misc{domyn_edge_teacher_math_mix_2026,
  title     = {Domyn Edge Teacher Math Mix},
  author    = {Domyn},
  year      = {2026},
  note      = {Technical report, forthcoming in November 2026.},
  publisher = {Hugging Face},
  url       = {https://huggingface.co/datasets/domyn/Domyn-Edge-Teacher-Math-Mix}
}
```