Medical question answering systems must balance usefulness with safety, particularly in low-resource linguistic settings where robustness is limited and hallucinations can cause harm. We present TRUMEDIQA, a reproducible multilingual medical QA pipeline for Moroccan Darija, Arabic, French, and English, deployed on WhatsApp with text and voice interactions. TRUMEDIQA uses layered decision-making: (i) language identification, (ii) a pre-retrieval intent router that maps queries to one of 38 clinical FAQ categories to constrain retrieval, and (iii) post-retrieval LLM-based re-ranking that selects the best candidate answer or returns a null decision to trigger a safe fallback (abstention). Answers are retrieved from a curated FAQ knowledge base validated by medical professionals. We evaluate TRUMEDIQA with 21 participants submitting 290 questions across four languages. An expert annotator labels each interaction as relevant, acceptable, or irrelevant, and we also measure correct abstentions when no suitable answer exists in the knowledge base. An ablation study shows that routing and re-ranking improve the weighted relevance score from 0.25 to 0.94 and precision from 0.53 to 0.98 versus a naïve retrieval baseline, while increasing correct abstention on unanswerable queries from 4.38% to 69.77%.

Despite recent advances in large language models (LLMs), most question-answering (QA) systems remain English-centric and poorly suited to domain-specific scientific texts. This linguistic and domain bias poses a major challenge in botany, where a substantial portion of knowledge is documented in French. We introduce F-LoRA-QA, a fine-tuned LLaMA-based pipeline for French botanical QA, leveraging Low-Rank Adaptation (LoRA) for efficient domain adaptation. We construct a specialized dataset of 16,962 question-answer pairs extracted from scientific flora descriptions and fine-tune LLaMA models to retrieve structured knowledge from unstructured botanical texts. Expert-based evaluation confirms the linguistic quality and domain relevance of generated answers. Compared to baseline LLaMA models, F-LoRA-QA achieves a 300% BLEU score increase, 70% ROUGE-1 F1 gain, +16.8% BERTScore F1, and Exact Match improvement from 2.01% to 23.57%. These results demonstrate the effectiveness of adapting LLMs to low-resource scientific domains and highlight the potential of our approach for automated trait extraction and biodiversity data structuring.

Large Language Models (LLMs) are increasingly used in scientific question answering (QA), including high-stakes fields such as biodiversity informatics. However, standard evaluation metrics such as BLEU, ROUGE, Exact Match (EM), and BERTScore remain poorly aligned with the factual and domain-specific requirements of these tasks. In this work, we investigate the gap between automatic metrics and expert judgment in botanical QA by comparing metric scores with human ratings across five dimensions: accuracy, completeness, relevance, fluency, and terminology usage. Our results show that standard metrics often misrepresent response quality, particularly in the presence of paraphrasing, omission, or domain-specific language. Through both quantitative analysis and qualitative examples, we show that high-scoring responses may still exhibit critical factual errors or omissions. These findings highlight the need for domain-aware evaluation frameworks that incorporate expert feedback and raise important ethical concerns about the deployment of LLMs in scientific contexts.