VietMed-MCQ: New Dataset and Benchmark for Vietnamese Traditional Medicine Evaluation with LLMs
Researchers introduced VietMed-MCQ, a dataset of 3,190 multiple-choice questions on Vietnamese Traditional Medicine generated using a consistency-filtered synthesis framework to address the scarcity of high-quality benchmarks in this specialized domain. The dataset underwent expert validation achieving 94.2% approval and was used to benchmark seven open-source language models. Results showed that general-purpose models with Chinese language priors outperformed Vietnamese-specific models, though all models struggled with complex diagnostic reasoning.
VietMed-MCQ addresses a significant gap in medical AI evaluation by providing the first large-scale, structured benchmark for Vietnamese Traditional Medicine. The dataset was created through a Retrieval-Augmented Generation (RAG) pipeline incorporating an automated consistency check mechanism with dual-model validation to ensure reasoning quality. The 3,190 questions span three difficulty levels and were validated by one medical expert and four students, achieving substantial inter-rater agreement (Fleiss' kappa = 0.82). Benchmarking of seven open-source models revealed that general-purpose models benefited from cross-lingual conceptual transfer from Chinese language training, while Vietnamese-centric models underperformed, suggesting that specialized cultural and linguistic knowledge alone does not guarantee superior performance in this domain. All tested models demonstrated limitations in handling complex diagnostic reasoning tasks.
What's missing
The authors acknowledge that their substring-based evidence checking mechanism has known limitations, though the specific nature and impact of these limitations on dataset quality are not detailed. Additionally, the study does not discuss potential biases in the expert validation process or how the four student validators' medical expertise level may have affected inter-rater agreement.
What different sources said
- arXiv cs.CLCenter
VietMed-MCQ: A Consistency-Filtered Data Synthesis Framework for Vietnamese Traditional Medicine Evaluation
Related
Genetic Drift, Not Selection, Drives Rapid Feather Color Evolution in Island Bird Radiation
A new study of an island bird radiation found that rapid evolution of feather coloration is driven primarily by genetic drift in small populations rather than sexual or ecological selection. The research integrated whole-genome data with detailed plumage measurements across complete species sampling to test whether signaling trait evolution correlates with speciation rates. The findings suggest that neutral demographic processes play a central role in generating phenotypic diversity during island radiations, challenging assumptions about the mechanisms driving rapid evolution.
New AI Model Improves Prediction of Therapeutic Peptide Function from Protein Sequences
Researchers developed a lightweight CNN classifier that predicts whether peptide sequences have therapeutic properties, trained on a database of 54,655 peptides across 48 functional categories. The model uses a novel negative sampling strategy to reduce false positive rates from over 60% in previous approaches to 2.1%. This advancement could accelerate drug discovery by enabling faster computational screening of peptide candidates before expensive experimental testing.
Study Shows Different Metabolic Stress Models Produce Distinct Effects on Human Neuronal Networks
Researchers tested three common in vitro metabolic stress models on human-derived neuronal networks and found each produced different patterns of neuronal activity and cell damage. The models tested were hypoxia alone, oxygen-glucose deprivation (OGD), and hypoxia combined with glutamate exposure. The findings suggest that choice of experimental model significantly affects results and that combining electrophysiological and structural analyses is important for accurately assessing metabolic stress in stroke research.