Update README.md

README.md

@@ -11,13 +11,24 @@ size_categories:
 - 10K<n<100K
 ---
 
+---
+pretty_name: MedInjection-FR — Native Subset
+language:
+- fr
+license: mit
+task_categories:
+- text-generation
+- question-answering
+size_categories:
+- 100K<n<300K
+---
+
 # MedInjection-FR — Native Subset 🇫🇷
 
 ## Summary
 
 The **Native** component of **MedInjection-FR** comprises **French biomedical instructions and question–answer pairs** natively written in French.  
-It forms the core of the dataset and reflects **authentic medical reasoning and linguistic formulations**, sourced from curated corpora such as *FrBMedQA*, *FrenchMedMCQA*, and *MediQAl*.
-
+It forms the core of the dataset and reflects **authentic medical reasoning and linguistic formulations**, sourced from curated corpora and educational materials.  
 This subset serves as the **high-quality reference supervision** for instruction tuning of large language models (LLMs) in French biomedical contexts.
 
 ## Motivation
@@ -27,34 +38,33 @@ The Native subset bridges this gap, providing instruction–response data derive
 
 ## Composition
 
-- **Languages:** French (`fr`)
-- **Domain:** Medicine, Clinical reasoning, Biology
-- **Task types:**  
-  - Multiple-choice (single and multiple answers)  
-  - Open-ended questions (diagnostic reasoning, treatment, translation-like reformulations)
-- **Structure:**
-Train/
-mcq/
-mcqu/
-oeq/
-Validation/
-mcq/
-mcqu/
-oeq/
-Test/
-mcq/
-mcqu/
-oeq/
-
-java
-Copy code
-Each JSON record includes:
-```json
-{
-  "instruction": "...",
-  "context": "...",        // optional
-  "options": ["..."],      // for MCQ/MCQU
-  "answer": "...",
-  "origin": "FrBMedQA",
-  "type": "mcqu"
-}
+The native component combines curated datasets and web-scraped French medical resources to reflect authentic domain knowledge. It integrates the following resources:
+
+- **S-Editions**~\cite{S-Editions}:  
+  526 question–answer pairs from a French educational platform for medical students.
+  
+- **MediQAl**~\cite{bazoge2025mediqal}:  
+  32 603 items from national medical examinations covering 41 medical specialties.
+
+- **FrenchMedMCQA**~\cite{labrak2023frenchmedmcqa}:  
+  3 105 pharmacy-focused multiple-choice questions.
+
+- **mlabonne/medical-cases-fr**~\cite{mlabonne_medical_cases_fr} and **mlabonne/medical-mcqa-fr**~\cite{mlabonne_medical_mqca_fr}:  
+  12 194 examples originating from French medical exam databases.
+
+- **FrBMedQA**~\cite{kaddari2022frbmedqa}:  
+  19 836 questions derived from French biomedical Wikipedia articles spanning eight UMLS semantic groups (chemicals and drugs, anatomy, physiology, disorders, phenomena, procedures, genes and molecular sequences, and devices).  
+  Originally closed-form, these questions were reformulated into multiple-choice format using *GPT-4o-mini*~\cite{hurst2024gpt} for standardization.
+
+- **Parallel Biomedical Translation Corpora**~\cite{biomedical_translation_corpora}:  
+  Bilingual biomedical translation data from the WMT challenge repositories were reformulated into instruction–response pairs.  
+  Each instruction requests the **French translation** of an English biomedical passage, reframing translation as an instruction-following task aligned with the native portion.
+
+## Use
+
+Used to fine-tune biomedical LLMs for:
+- Question answering and clinical reasoning  
+- Instruction-following in French  
+- Evaluating cross-domain instruction generalization
+
+