cs.CLNov 16, 2026

Medical question answering collapses in Hausa: a pre-registered pilot on three small open models

Authors: Abbas Sani

Organizations: Mewar University

Abstract

A system that reads nothing and always answers the most frequent letter scores \BaselineP{} on our Hausa medical multiple-choice set. \NBelowBaseline{} of the \NModels{} small open models we evaluate do not beat it. We lead with that because the usual summary of a multilingual result -- a percentage-point drop from English -- does not say whether what survives the drop is performance at all. On a paired English--Hausa set of \NEval{} MedMCQA items, three instruction-tuned models of 3--4B parameters lose between \GapMedGemma{} and \GapQwen{} points of accuracy, every gap significant after Holm correction. Back-translating the Hausa items into English recovers between half and two-thirds of the gap for the two general-purpose models but not demonstrably for the medically adapted one, so the deficit is substantially, though not wholly, a matter of the language of presentation. Answers are also far less stable: agreement across three option orderings falls from \ConsEnQwen{} to \ConsHaQwen{} in the worst case. An exploratory error analysis of \ENRows{} discordant items attributes most of the damage to medical vocabulary lost in translation (\ETerm{}, 95% CI \ETermLo{}--\ETermHi{}), usually in the question stem. A fourth registered hypothesis proposed closing the gap with self-generated preference pairs. Its pilot gate found nothing to train on: asked to reason in Hausa rather than to pick a letter, the model echoes the prompt back, switches into Swahili, or simply stops without answering; only \GateParse{} of chains are parseable. We report that intervention as not run, which is itself the sharper result -- a forced-choice benchmark cannot see a model that has no generative command of the language at all. All hypotheses, thresholds and decision rules were registered before any model was run, and all deviations are logged.

Figures & tables

Explore similar work

Nov 12, 2026

Calibration and Conformal Prediction for Medical Question Answering in Arabic and English: Evidence from Two Non-Parallel Benchmarks

Large language models (LLMs) are increasingly tested on medical exam questions, but accuracy alone does not show when an answer can be trusted. I study the calibration and uncertainty of two open-weight models, Qwen2.5-7B-Instruct and Jais-adapted-7B-chat (both run in 4-bit precision), on 2,816 English multiple-choice questions from MedMCQA and 4,927 Arabic multiple-choice questions from MedAraBench. Accuracy was 0.561 (English) and 0.430 (Arabic) for Qwen, and 0.369 and 0.308 for Jais. Raw option probabilities were overconfident in all four settings (expected calibration error 0.242 to 0.486), and a single temperature parameter removed 82% to 93% of that error. Split conformal prediction reached its nominal coverage in all 24 settings I tested (within 0.006), but the sets were large when accuracy was low: at 90% coverage, temperature-scaled sets contained on average 63% (Qwen, English), 78% (Qwen, Arabic), 86% (Jais, English) and 89% (Jais, Arabic) of the options. The margin between the top two option probabilities separated correct from incorrect answers best for Qwen on English (AUROC 0.734) and worst for Jais on Arabic (0.579). Answering only the 20% most confident questions lowered Qwen's English error from 0.439 to 0.098. Jais showed a strong prior for option C (49% of English and 60% of Arabic predictions). Scoring Latin letters only cut the Arabic share to 29% but left accuracy, raw ECE and AUROC nearly unchanged. Because the two benchmarks differ in subject, level and number of options, I do not attribute differences between languages to language itself. Code is available at https://github.com/Mariyaaleem/arabic-english-medqa-uncertainty.
Jul 19, 2026

Safety That Does Not Transfer: Cross-Lingual Clinical Correctness Drift in Deployable Medical Language Models

Safety evaluation of large language models is conducted predominantly in English and predominantly on frontier systems. Neither condition describes how such models are encountered in low-resource health settings, where small quantised systems are run locally and queried in local languages. We ask whether clinical safety established in English transfers to Hausa, and whether any failure is attributable to the language, the clinical task, or the class of model that low-resource deployment admits. Matched English-Hausa question pairs were built for three conditions of high burden in northern Nigeria: malaria, sickle cell disease, and tuberculosis, probing knowledge recall, emergency triage, a leading question inviting a contraindicated action, and a traditional-remedy claim. Six models were evaluated: five locally deployable systems of 4-9 billion parameters, two medically fine-tuned, and one frontier system. All 128 responses were scored against Nigerian national treatment guidelines by two fluent Hausa speakers working independently and blind to one another. Among locally deployable models, mean clinical correctness fell from 1.57 in English to -0.03 in Hausa, on a scale where 2 denotes a correct answer and -1 an actively harmful one. The frontier model moved from 2.00 to 1.75 and produced no response judged harmful in either language. Drift was consistent across all three conditions. Inter-rater agreement was substantial for clinical correctness (kappa = 0.70); agreement on harm was initially poor (kappa = 0.22) and is examined in detail. Because a frontier model answers the same questions competently in Hausa, the deficit is a property neither of the language nor of the clinical material, but of the deployable tier.
Aug 20, 2026

HealMed: Multilingual Evaluation of Large Language Models in Medicine

We present HealMed, an expert-reviewed benchmark for multilingual evaluation of large language models in medicine. HealMed contains 1,000 examples in each of nine languages, drawn from nine datasets and covering three task formats: MCQA, NLI and open-ended QA. The benchmark was developed over two years by 23 physicians and medical experts based across nine countries and regions. Each translation was evaluated and revised by two experts fluent in English and the corresponding target language. On HealMed, performance declined most in low-resource languages, although the size of the gap varied markedly across languages and models. The strongest proprietary models were the most stable across languages, whereas many open-source and medically specialized models showed larger and less consistent gaps. Medical specialization alone did not ensure multilingual robustness. Furthermore, expert revision could either raise or lower measured performance, indicating that translation quality materially affects cross-language evaluation results.