--- license: mit pretty_name: Multilingual GSM-Symbolic tags: - math - reasoning - symbolic - multilingual # GENERATED METADATA START (updated on every release) language: - afr - amh - ara - arz - asm - azb - bam - bar - bel - ben - bho - bos - bul - cat - ceb - ces - ctg - dan - deu - dyu - ell - eng - est - fin - fra - fuv - gaz - gle - guj - hat - hau - hin - hne - hrv - hun - ibo - ind - isl - ita - jav - jpn - kan - kaz - khm - kin - kmr - kor - ktu - lav - lin - lit - lug - mag - mai - mal - mar - mlt - mos - mya - nld - nno - nob - npi - nso - nya - ory - pan - pbu - pcm - pes - pnb - pol - por - ron - run - rus - sck - sin - skr - slk - slv - sna - snd - som - spa - sun - swe - swh - tam - tel - tgl - tha - tsn - tur - uig - ukr - urd - uzn - vie - vjk - wol - xho - yor - yue - zho - zlm - zul size_categories: - 100K [![GitHub](https://img.shields.io/badge/GitHub-multilingual--gsm--symbolic-blue?logo=github)](https://github.com/centre-for-humanities-computing/multilingual-gsm-symbolic) [![Version](https://img.shields.io/badge/version-v0.5.5-green)](https://huggingface.co/datasets/danish-foundation-models/multilingual-gsm-symbolic/tree/v0.5.5) | | | |---|---| | **Version** | [`v0.5.5`](https://huggingface.co/datasets/danish-foundation-models/multilingual-gsm-symbolic/tree/v0.5.5) | | **Released** | 2026-10-06 | | **Languages** | 108 (19 validated) | **Multilingual GSM-Symbolic** is a benchmark for evaluating arithmetic reasoning in large language models across many languages. It extends Apple's [GSM-Symbolic](https://machinelearning.apple.com/research/gsm-symbolic) approach by providing symbolic templates that generate thousands of structurally equivalent but numerically distinct math problems. Templates and generation are handled by the [`multilingual-gsm-symbolic`](https://github.com/centre-for-humanities-computing/multilingual-gsm-symbolic) package. The dataset lets you test whether a model genuinely understands a problem or merely pattern-matches on the specific numbers it saw during training. Comparing the `original` and `synthetic` splits directly measures this gap.
A symbolic template is translated into several languages, rendered into many concrete problems, and used to measure the gap between languages and between original and synthetic problems
## Loading the dataset ### With πŸ€— datasets ```python from datasets import load_dataset # English synthetic test split, pinned to a version for reproducible results (see the version table at the top) eng = load_dataset("danish-foundation-models/multilingual-gsm-symbolic", name="eng", split="test_synthetic", revision="v0.5.0") # Original test split of a language that has not yet been validated swh = load_dataset("danish-foundation-models/multilingual-gsm-symbolic", name="swh", split="test_original_unvalidated", revision="v0.5.0") ``` ### With inspect-ai You can evaluate with [inspect-ai](https://inspect.aisi.org.uk) using the `eval.yaml` in this repository: ```bash inspect eval hf/danish-foundation-models/multilingual-gsm-symbolic/synthetic_eng --model openai/gpt-5.4-nano --reasoning-effort low ``` To pin a version, add it after the task id: `hf/danish-foundation-models/multilingual-gsm-symbolic/synthetic_eng@v0.5.0`. Task ids are `original_` and `synthetic_` for validated languages (e.g. `original_deu`) and `original_unvalidated_` and `synthetic_unvalidated_` for the rest. Omit the task id to run all of them. Each task prompts the model in the language of the problem, asking it to solve the problem step by step and put the final answer in `\boxed{}`. The instructions are part of the templates in the GitHub repository (`instruction.toml` in each language folder) and can be loaded with `multilingual_gsm_symbolic.load_instruction`. Most instructions are machine translated and are annotated as validated once checked by a native speaker. ## Dataset structure Each language is a separate **config** (subset) named by its three-letter language code (e.g. `eng`, `dan`). `eng_metric` is a variant of English using metric units. | Split | Description | |-------|-------------| | `test_original` | The concrete GSM problems for that language (100 for validated languages) | | `test_synthetic` | 20 generated variants per template (2,000 problems for validated languages) | | `test_original_unvalidated`, `test_synthetic_unvalidated` | The same splits for languages that have not yet been validated by a native speaker. Templates that could not be translated reliably are left out, so some languages have fewer than 100 | | `train_original` | English only: 191 concrete training problems from GSM8K | | `train_synthetic` | English only: 20 generated variants per training template (3,820 problems). The training templates are machine-generated and not validated | ### Fields | Field | Type | Description | |-------|------|-------------| | `question` | string | The math problem | | `answer` | string | Step-by-step solution ending with `#### ` | | `target` | string | The final numeric answer (extracted from `answer`) | | `language` | string | Three-letter language code (e.g. `eng`, `dan`) | | `source_id` | int | Problem index in the original GSM8K dataset | ### Answer format Answers follow the GSM8K convention β€” reasoning steps followed by a final numeric answer: ``` At 3 miles/hour, it will take 42/3=14 hours for the fog to cover the city. #### 14 ``` ## Languages The dataset contains 19 validated languages, where every template has been computationally validated and checked by native speakers: | Language | Computationally validated | Human validated | Error analysis | | --- | --- | --- | --- | | `ara` | βœ“ | by a native speaker | βœ“ | | `dan` | βœ“ | by three native speakers | βœ“ | | `deu` | βœ“ | by two native speakers | βœ“ | | `eng` | βœ“ | by a native speaker | βœ“ | | `eng_metric` | βœ“ | by a native speaker | | | `est` | βœ“ | by a native speaker | βœ“ | | `fra` | βœ“ | by a native speaker | βœ“ | | `hin` | βœ“ | by a native speaker | βœ“ | | `isl` | βœ“ | by native speakers | βœ“ | | `ita` | βœ“ | by two native speakers | βœ“ | | `jpn` | βœ“ | by a native speaker | βœ“ | | `mar` | βœ“ | by a native speaker | βœ“ | | `nld` | βœ“ | by a native speaker | βœ“ | | `pan` | βœ“ | by two native speakers | βœ“ | | `rus` | βœ“ | by a native speaker | βœ“ | | `swe` | βœ“ | by 5 native speakers | βœ“ | | `ukr` | βœ“ | by a native speaker | βœ“ | | `urd` | βœ“ | by a native speaker | βœ“ | | `zho` | βœ“ | by a native speaker | βœ“ | The remaining 89 languages are machine translated and only computationally validated. Their splits are suffixed with `_unvalidated`: `afr`, `amh`, `arz`, `asm`, `azb`, `bam`, `bar`, `bel`, `ben`, `bho`, `bos`, `bul`, `cat`, `ceb`, `ces`, `ctg`, `dyu`, `ell`, `fin`, `fuv`, `gaz`, `gle`, `guj`, `hat`, `hau`, `hne`, `hrv`, `hun`, `ibo`, `ind`, `jav`, `kan`, `kaz`, `khm`, `kin`, `kmr`, `kor`, `ktu`, `lav`, `lin`, `lit`, `lug`, `mag`, `mai`, `mal`, `mlt`, `mos`, `mya`, `nno`, `nob`, `npi`, `nso`, `nya`, `ory`, `pbu`, `pcm`, `pes`, `pnb`, `pol`, `por`, `ron`, `run`, `sck`, `sin`, `skr`, `slk`, `slv`, `sna`, `snd`, `som`, `spa`, `sun`, `swh`, `tam`, `tel`, `tgl`, `tha`, `tsn`, `tur`, `uig`, `uzn`, `vie`, `vjk`, `wol`, `xho`, `yor`, `yue`, `zlm`, `zul`. The evaluation instructions (see `eval.yaml`) have been checked by native speakers for: `ara`, `dan`, `deu`, `eng`, `eng_metric`, `est`, `fra`, `hin`, `isl`, `ita`, `jpn`, `mar`, `nld`, `pan`, `rus`, `swe`, `ukr`, `urd`, `zho`. Want to add or validate a language? It only requires validating 100 templates. Read more about how to contribute [here](https://github.com/centre-for-humanities-computing/multilingual-gsm-symbolic#want-to-add-a-new-language). ## Versions The dataset is generated from the templates in the [GitHub repository](https://github.com/centre-for-humanities-computing/multilingual-gsm-symbolic) and published automatically with every release of the `multilingual-gsm-symbolic` package. Each release is tagged here with the same version (e.g. `v0.5.0`). The data changes between versions when templates are fixed or languages are validated, so pin the version to make your results reproducible (see [Loading the dataset](#loading-the-dataset)). ## Example analysis As an example of what the dataset can be used for, we evaluated `openai/gpt-5.4-nano` (reasoning effort: low) with [inspect-ai](https://inspect.aisi.org.uk) on English and Danish. These results were produced in April 2026 on an earlier version of the dataset, so they are not directly comparable to evaluations on later versions. | Language | Original accuracy | Synthetic accuracy | |----------|------------------|--------------------| | English | 90.0% | 75.2% | | Danish | 83.2% | 70.2% | The gap between `original` and `synthetic` accuracy reflects performance degradation on novel number combinations β€” a proxy for how much a model relies on memorisation vs. genuine reasoning. ### Performance degradation (original β†’ synthetic)
Distribution of synthetic set accuracy
### Language gap (English vs. Danish, synthetic)
Overlapping accuracy distributions for English and Danish
### Accuracy by problem difficulty (reasoning steps)
Accuracy as a function of the number of reasoning steps
Eval logs are available in the [`logs/`](https://huggingface.co/datasets/danish-foundation-models/multilingual-gsm-symbolic/tree/main/logs) folder of this repository. ## Citation If you use this dataset, please cite the [Multilingual GSM-Symbolic paper](https://arxiv.org/abs/2610.03367): ```bibtex @misc{enevoldsen2026multilingualgsmsymbolicdeterminescapability, title={Multilingual GSM-Symbolic: What determines capability transfer across languages?}, author={Kenneth Enevoldsen and Riley Herchert and Sofie Mosegaard and Dan Saattrup Smart and Simon Enni and Isaac Chung and Sofie Bruun and Ayush Sunil Munot and Max MΓΌller-Eberstein and Adnan El-Assadi and Elisa Bassignana and Gianluca Barmina and Hafsteinn Einarsson and Iben Nyholm Debess and Linda Freienthal and Lukas Galke Poech and Mike Zhang and Nicolas Legrand and Vladimir Salnikov and Yevhen Kostiuk and Zafar Hussain and Sagandeep Kaur and Agnes ToftgΓ₯rd and Marie Mattson and Kristoffer Nielbo}, year={2026}, eprint={2610.03367}, archivePrefix={arXiv}, primaryClass={cs.AI}, url={https://arxiv.org/abs/2610.03367}, } ``` ---

A Danish Foundation Models dataset