--- license: mit task_categories: - question-answering - table-question-answering tags: - smoldataenvs - data-analysis - agent - rl-environment - reinforcement-learning - code-agent configs: - config_name: default data_files: - split: train path: data/train-* - split: test path: data/test-* - split: eval path: data/eval-* ---
SmolDataEnvs # 📈 SmolDataEnvs [![Collection](https://img.shields.io/badge/%F0%9F%A4%97%20Collection-SmolDataEnvs-FFD21E?style=for-the-badge&labelColor=1a1a1a)](https://huggingface.co/collections/FineEnvs/smoldataenvs)
> **5.5K+ RL tasks for hill-climbing small models in code and data science.**
Reward and held-out pass@k climbing over 1,119 GRPO steps A 2B model on these tasks. Left: what it optimises. Right: 144 held-out tasks it never trains on.
Two runs over the same 5,000 tasks: shuffled against a curriculum ordered easiest to hardest.
Data-analysis tasks as a plain, load-and-go dataset: **no runtime, no framework required**. Each row is one self-contained task: a real tabular dataset, a question about it, and a gold answer a bundled grader can check deterministically. Load it, prompt any model however you like, grade the result. This is the front door. If you want the tasks as runnable sandboxed environments, use the [Harbor suites](https://huggingface.co/datasets/FineEnvs/SmolDataEnvs-harbor-train); if you want demonstrations to fine-tune on, use [`-sft`](https://huggingface.co/datasets/FineEnvs/SmolDataEnvs-sft). ## Splits | Split | Tasks | Easy | Medium | Hard | What it's for | |---|---|---|---|---|---| | `train` | 5,000 | 1,433 | 2,845 | 722 | training | | `test` | 250 | 33 | 118 | 99 | held-out benchmark, deliberately harder | | `eval` | 144 | 16 | 74 | 54 | quick validation during a run | The held-out splits are harder than train by construction: train is 29% easy and 14% hard, the held-out splits are 11–13% easy and 38–40% hard. Worth knowing before you read any eval number. ## What's in a row | Column | Meaning | |---|---| | `task_id`, `source_row_id` | identifiers | | `question` | the question to answer | | `answer` | the gold answer | | `reward_mode`, `atol`, `rtol` | how to grade it: match type and numeric tolerances | | `difficulty_level` (1–5), `difficulty_tier` | difficulty | | `kaggle_dataset` | the source dataset | | `hf_bucket`, `bucket_prefix`, `files` | where the input files live and what they are | | `instruction` | the full agent prompt | | `package_tier` | environment sizing hint | ## Load it ```python from datasets import load_dataset ds = load_dataset("FineEnvs/SmolDataEnvs", split="test") row = ds[0] print(row["question"], "→", row["answer"], f"({row['reward_mode']})") ``` ## Grab the data files for a task The tables live in a Hugging Face **bucket**, so they come down with the bucket API rather than `snapshot_download`: ```python from huggingface_hub import list_bucket_tree, download_bucket_files prefix = row["bucket_prefix"].rstrip("/") + "/" items = [i for i in list_bucket_tree(row["hf_bucket"], prefix=prefix, recursive=True) if getattr(i, "type", None) == "file"] download_bucket_files(row["hf_bucket"], files=[(i.path, "input/" + i.path.split("/")[-1]) for i in items]) ``` ## Grade a prediction `grader.py` ships in this repo. It scores an answer through a ladder of checks: exact → numeric with `atol`/`rtol` → list and percent normalisation → symbolic equivalence: ```python from huggingface_hub import hf_hub_download import importlib.util, sys path = hf_hub_download("FineEnvs/SmolDataEnvs", "grader.py", repo_type="dataset") spec = importlib.util.spec_from_file_location("grader", path) grader = importlib.util.module_from_spec(spec) sys.modules["grader"] = grader # the dataclasses inside it need this spec.loader.exec_module(grader) r = grader.grade(row["answer"], my_prediction, reward_mode=row["reward_mode"], abs_tol=row["atol"], rel_tol=row["rtol"]) print(r.reward, r.method) # 1.0 exact | 0.0 miss ``` ## Where it comes from Built from the [jupyter-agent dataset](https://huggingface.co/datasets/jupyter-agent/jupyter-agent-dataset), real data-science notebooks over 471 Kaggle datasets. Every question–answer pair was extracted and then **verified**: strong agent models had to solve the task in a live sandbox and reproduce the gold answer under deterministic grading. Anything ambiguous or un-checkable was dropped. So every task here is known-solvable and unambiguously gradable. **Verified by a checker, not judged by a model.** Grading is an exact comparison against a known answer, through a ladder of checks: exact match → numeric with tolerances → list and percent normalisation → symbolic equivalence. No LLM sits in the reward path, so the signal does not drift when you change the grader's model, because there isn't one. ## The family | Repo | What it is | |---|---| | [`SmolDataEnvs`](https://huggingface.co/datasets/FineEnvs/SmolDataEnvs) | the tasks as plain rows, load it and prompt any model | | [`SmolDataEnvs-sft`](https://huggingface.co/datasets/FineEnvs/SmolDataEnvs-sft) | 4,677 verified agent trajectories, TRL-ready | | [`SmolDataEnvs-harbor-train`](https://huggingface.co/datasets/FineEnvs/SmolDataEnvs-harbor-train) | 5,000 tasks as Harbor environments | | [`SmolDataEnvs-harbor-test`](https://huggingface.co/datasets/FineEnvs/SmolDataEnvs-harbor-test) | 250 held-out, deliberately harder | | [`SmolDataEnvs-harbor-eval`](https://huggingface.co/datasets/FineEnvs/SmolDataEnvs-harbor-eval) | 144 for quick validation during a run | ## Train on it The simplest path, a notebook and a single-file script you can hand to HF Jobs, lives in [FineEnvs/04-smoldataenvs](https://github.com/adithya-s-k/FineEnvs/tree/main/04-smoldataenvs). ## Citation ```bibtex @misc{fineenvs, author = {Kolavi, Adithya S}, title = {FineEnvs: Open Source RL Environments for LLM Agents}, year = {2026}, url = {https://github.com/adithya-s-k/FineEnvs} } ```