arXiv:2603.03680v1 Announce Type: new Abstract: Large Language Model (LLM) agents have demonstrated remarkable proficiency in learned tasks, yet they often struggle to adapt to non-stationary environments with feedback. While In-Context Learning and external memory offer some flexibility, they fail to internalize the adaptive ability required for long-term improvement. Meta-Reinforcement Learning (meta-RL) provides an alternative by embedding the learning process directly within the model. However, existing meta-RL approaches for LLMs focus primarily on exploration in single-agent settings, neglecting the strategic exploitation necessary for multi-agent environments. We propose MAGE, a meta-RL framework that empowers LLM agents for strategic exploration and exploitation. MAGE utilizes a multi-episode training regime where interaction histories and reflections are integrated into the context window. By using the final episode reward as the objective, MAGE incentivizes the agent to refine its strategy based on past experiences. We further combine population-based training with an agent-specific advantage normalization technique to enrich agent diversity and ensure stable learning. Experiment results show that MAGE outperforms existing baselines in both exploration and exploitation tasks. Furthermore, MAGE exhibits strong generalization to unseen opponents, suggesting it has internalized the ability for strategic exploration and exploitation. Code is available at https://github.com/Lu-Yang666/MAGE.

arXiv:2603.02221v1 Announce Type: cross Abstract: In healthcare tabular predictions, classical models with feature engineering often outperform neural approaches. Recent advances in Large Language Models enable the integration of domain knowledge into feature engineering, offering a promising direction. However, existing approaches typically rely on a broad search over predefined transformations, overlooking downstream model characteristics and feature importance signals. We present MedFeat, a feedback-driven and model-aware feature engineering framework that leverages LLM reasoning with domain knowledge and provides feature explanations based on SHAP values while tracking successful and failed proposals to guide feature discovery. By incorporating model awareness, MedFeat prioritizes informative signals that are difficult for the downstream model to learn directly due to its characteristics. Across a broad range of clinical prediction tasks, MedFeat achieves stable improvements over various baselines and discovers clinically meaningful features that generalize under distribution shift, demonstrating robustness across years and from ICU cohorts to general hospitalized patients, thereby offering insights into real-world deployment. Code required to reproduce our experiments will be released, subject to dataset agreements and institutional policies.

arXiv:2602.17686v2 Announce Type: replace-cross Abstract: Distilling Chain-of-Thought (CoT) reasoning from large language models into compact student models presents a fundamental challenge: teacher rationales are often too verbose for smaller models to faithfully reproduce. Existing approaches either compress reasoning into single-step, losing the interpretability that makes CoT valuable. We present a three-stage curriculum learning framework that addresses this capacity mismatch through progressive skill acquisition. First, we establish structural understanding via masked shuffled reconstruction. Second, we apply Group Relative Policy Optimization (GRPO) on masked completion tasks, enabling the model to discover its own balance between accuracy and brevity. Third, we identify persistent failure cases and guide the student to internalize teacher knowledge through targeted rewriting, again optimized with GRPO. Experiments on GSM8K demonstrate that our approach enables Qwen2.5-3B-Base to achieve an 11.29 percent accuracy improvement while reducing output length by 27.4 percent, surpassing both instruction-tuned variants and prior distillation methods.