Update README.md

README.md

@@ -1,4 +1,27 @@
 ---
+pretty_name: Exo7MCQ
+language:
+- fr
+license: cc-by-nc-sa-4.0
+task_categories:
+- question-answering
+- multiple-choice
+size_categories:
+- n<1K
+source_datasets:
+- original
+annotations_creators:
+- found
+tags:
+- math
+- education
+- french
+- mcq
+- multi-label
+- benchmark
+- undergraduate
+- exo7
+- lighteval
 dataset_info:
   features:
   - name: id
@@ -27,3 +50,128 @@ configs:
   - split: test
     path: data/test-*
 ---
+
+# Exo7MCQ
+
+**Exo7MCQ** is a native-French **multi-label** multiple-choice benchmark for
+undergraduate mathematics: **946 questions** spanning L1–L3 topics (logic and set theory,
+analysis, linear algebra, arithmetic, differential equations, probability), extracted from
+the LaTeX question banks of [Exo7](http://exo7.emath.fr/).
+
+Unlike a standard MCQ set, **each option carries its own correctness label**: a question may
+have one, two, three or four correct options among its 2–6 choices. Answering it is a
+per-option binary decision, not a "pick one of four".
+
+## Quick start
+
+```python
+from datasets import load_dataset
+
+ds = load_dataset("cea-list-ia/Exo7MCQ", split="test")
+
+row = ds[0]
+row["question"]                        # the stem
+row["targets"]["choices"]              # the options
+row["targets"]["labels"]               # parallel 0/1 vector, 1 = correct
+```
+
+## Dataset structure
+
+### Fields
+
+| field | type | description |
+|---|---|---|
+| `id` | `int64` | Stable identifier, `0`–`954`. **Non-contiguous**: ids are indices into the 955-item pre-filter set, and 9 items were removed (see *Creation*). |
+| `question` | `string` | Question stem in French with LaTeX (`$…$` and `\(…\)`). May be prefixed with shared context from its LaTeX subsection. |
+| `targets.choices` | `list[string]` | Answer options, 2–6 per item. |
+| `targets.labels` | `list[int64]` | Parallel multi-hot vector, `1` = correct, `0` = incorrect. |
+| `explanation` | `string` or `null` | Reference explanation from the source `explanations` block. Absent for 19 % of items. |
+| `category` | `string` | Source chapter title, underscored, 28 distinct values (e.g. `Equations_différentielles`, `Logique_--_Raisonnement_\|_100`). Some retain the original chapter number, so naming is inconsistent across banks. |
+
+
+### Example
+
+```json
+{
+  "id": 0,
+  "question": "Quelles sont les affirmations vraies ?",
+  "targets": {
+    "choices": [
+      "$x^3$ est une primitive de $3x^2+3$.",
+      "$x^3+3$ est une primitive de $3x^2$.",
+      "$\\ln(x^2+1)$ est une primitive de $\\frac 1{x^2+1}$.",
+      "$\\sqrt x$ est une primitive de $\\frac 1{2\\sqrt x}$ (sur $]0,+\\infty[$)."
+    ],
+    "labels": [0, 1, 0, 1]
+  },
+  "explanation": "Pour vérifier si une fonction \\(f\\) est une primitive d'une fonction \\(g\\), on calcule la dérivée de \\(f\\)…",
+  "category": "Equations_différentielles"
+}
+```
+
+
+## Evaluation
+
+Evaluated with [lighteval](https://github.com/huggingface/lighteval) via a community task
+module (`community_tasks/exo7.py`)
+
+## Dataset creation
+
+### Source data
+
+[Exo7](http://exo7.emath.fr/) is an open French undergraduate mathematics resource authored
+by university lecturers (Arnaud Bodin, Abdellah Hanani, Mohamed Mzari and others, published
+via Unisciel) and distributed as LaTeX sources. Five curated MCQ banks were used:
+
+| bank | topics | question blocks |
+|---|---|---:|
+| `arnaud` | general revision QCM | 30 |
+| `hilisit` | analysis, algebra | 163 |
+| `lille-1` | logic, sets, reals, sequences, limits, continuity, complex numbers, geometry | 360 |
+| `lille-2` | linear algebra, matrices, integrals, expansions | 360 |
+| `worms` | probability | 46 |
+
+
+### Extraction and filtering
+![exo7-pipeline](exo7_pipeline_en-1.png)
+
+
+## Licensing
+
+Exo7 materials are published under [**CC BY-NC-SA 4.0 FR**](https://creativecommons.org/licenses/by-nc-sa/4.0/deed.fr).
+This dataset is a derivative and inherits those terms.
+
+## Acknowledgements
+
+This work was granted access to the HPC resources of IDRIS under the allocation Grant AD011016978 made by GENCI.
+
+This publication was made possible by the use of the FactoryIA supercomputer, financially supported by the Ile-de-France Regional Council.
+
+We would also like to thank the [Banque Publique d'Investissement (BPI)](https://www.bpifrance.fr/) and the [OpenLLM consortium](https://openllm-france.fr/).
+
+## Citation
+
+```bibtex
+@inproceedings{duignan:hal-05708459,
+  TITLE = {{Jeux de donn{\'e}es en fran{\c c}ais pour l'affinage et l'{\'e}valuation de mod{\`e}les de langue g{\'e}n{\'e}ratifs dans le domaine des math{\'e}matiques}},
+  AUTHOR = {Duignan, Liam and Graiess, Asma and van Ypersele, Matteo and Deshayes-Chossart, J{\'e}r{\^o}me and Ferret, Olivier},
+  URL = {https://hal.science/hal-05708459},
+  BOOKTITLE = {{Actes de l'atelier Intelligence Artificielle et {\'E}DUcation (IA EDU)}},
+  ADDRESS = {Nantes, France},
+  EDITOR = {Badache, Ismail},
+  PUBLISHER = {{ATALA}},
+  PAGES = {74-97},
+  YEAR = {2026},
+  MONTH = Jun,
+  KEYWORDS = {Benchmark natif en fran{\c c}ais ; {\'E}valuation des LLMs en math{\'e}matiques},
+  PDF = {https://hal.science/hal-05708459v1/file/IA_EDU_5_2026.pdf},
+  HAL_ID = {hal-05708459},
+  HAL_VERSION = {v1},
+}
+
+```
+
+Please also credit the upstream source:
+
+> Exo7 — Bodin A., Blanc-Centi L., Borne N., Boutin B., Desideri L., Mégy D., Romon P.,
+> <http://exo7.emath.fr/> (CC BY-NC-SA).