The field of clinical decision support is witnessing significant advancements with the integration of large language models (LLMs). Recent developments focus on enhancing the collaborativeness and robustness of LLMs in medical decision-making scenarios. Researchers are exploring innovative methodologies to boost the performance of LLMs, including adaptive cluster collaborativeness and geometry-aware evaluation frameworks. These approaches aim to address the limitations of existing architectures, such as the lack of explicit component selection rules and the reliance on predefined LLM clusters. Furthermore, there is a growing emphasis on identifying and mitigating implicit biases in medical LLMs, as well as developing frameworks that can systematically reveal complex bias patterns. Noteworthy papers in this area include:

• A study that proposes an adaptive cluster collaborativeness methodology to enhance LLMs' medical decision support capacity, achieving state-of-the-art results on specialized medical datasets.
• A framework that combines knowledge graphs with auxiliary LLMs to reveal implicit biases in medical LLMs, demonstrating greater ability and scalability compared to other baselines.
• A geometry-aware evaluation framework that probes the latent robustness of clinical LLMs under structured adversarial edits, highlighting the importance of geometry-aware auditing in safety-critical clinical AI.