{"slug": "improving-heart-focused-medical-question-answering-in-llms-via-variance-aware", "title": "Improving Heart-Focused Medical Question Answering in LLMs via Variance-Aware Rubric Rewards with GRPO", "summary": "Researchers developed a variance-aware reward framework using Group Relative Policy Optimization (GRPO) to improve heart-focused medical question answering in large language models. The approach, which replaces traditional binary scoring with continuous analytical reward functions derived from rubric-based supervision, boosted accuracy from 0.362 to 0.502 and F1 from 0.532 to 0.668 on a heart-related subset of HealthBench. The findings demonstrate that rubric-based rewards can effectively enhance smaller models' medical reasoning performance while remaining competitive with larger models like GPT-OSS-120B.", "body_md": "arXiv:2606.05174v1 Announce Type: new\nAbstract: Large Language Models (LLMs) have shown strong promise in healthcare applications. Yet deploying general-purpose models in real-world settings remains difficult due to data privacy constraints, inference costs, and limited suitability for edge or on-device use. These challenges motivate the development of smaller, more efficient models that require robust post-training strategies to ensure reliable medical reasoning. In this work, we investigate Group Relative Policy Optimization (GRPO) for post-training LLMs on heart-focused medical question answering with rubric-based supervision derived from RaR-Medicine. We propose a Variance-Aware Reward Framework that extends the Explicit Aggregation and Implicit Aggregation strategies of Rubrics as Rewards by replacing weighted binary criterion aggregation and single overall Likert-style scoring with continuous analytical reward functions derived from criterion-level rubric outcomes. This formulation provides richer optimization signals for feedback that is sparse, multi-criteria, and difficult to verify automatically, and enables more stable on-policy reinforcement learning. On a held-out heart-related subset of HealthBench, our best GRPO variant improves accuracy from 0.362 to 0.502 and F1 from 0.532 to 0.668 relative to the Qwen3-14B base model, while remaining competitive with GPT-OSS-120B (0.508 accuracy, 0.674 F1). Our findings show that carefully designed rubric-based rewards provide a practical strategy for improving heart-focused medical question answering in LLMs, with potential to extend to other rubric-based tasks.", "url": "https://wpnews.pro/news/improving-heart-focused-medical-question-answering-in-llms-via-variance-aware", "canonical_source": "https://arxiv.org/abs/2606.05174", "published_at": "2026-06-05 04:00:00+00:00", "updated_at": "2026-06-05 04:19:52.055519+00:00", "lang": "en", "topics": ["large-language-models", "artificial-intelligence", "machine-learning", "natural-language-processing", "ai-research"], "entities": ["GRPO", "RaR-Medicine", "HealthBench", "Qwen3-14B", "GPT-OSS-120B"], "alternates": {"html": "https://wpnews.pro/news/improving-heart-focused-medical-question-answering-in-llms-via-variance-aware", "markdown": "https://wpnews.pro/news/improving-heart-focused-medical-question-answering-in-llms-via-variance-aware.md", "text": "https://wpnews.pro/news/improving-heart-focused-medical-question-answering-in-llms-via-variance-aware.txt", "jsonld": "https://wpnews.pro/news/improving-heart-focused-medical-question-answering-in-llms-via-variance-aware.jsonld"}}