Chain-of-thought prompting improves reasoning by generating intermediate steps as discrete text, but each step is forced to commit to a single token even when the model is uncertain. Continuous latent reasoning offers an alternative: a latent state can preserve multiple candidate interpretations without immediately selecting one token. Existing training-free approaches either place latent states only before decoding or switch between explicit and latent modes without optimizing the inserted state itself. We propose MidThink, a training-free framework that inserts and optimizes latent blocks both at the start of reasoning and adaptively during generation. When a confidence trigger detects sustained mid-stream uncertainty, MidThink inserts a new latent block and refines it with a rollout-based reward computed from short future continuations. We further show that initialization strongly affects whether a mid-stream block preserves a superposition over multiple tokens or collapses toward a single token. Across five mathematical reasoning benchmarks, MidThink improves over training-free latent-reasoning baselines, with the largest gains on the AIME benchmarks. Mechanistic analysis links these gains to three measurable properties of the latent block: distributional entropy across layers, attention centroid position, and stability under reinforcement learning updates.

Recent large language models are increasingly being adapted for medical reasoning, yet most open medical LLMs remain optimized for English-centric clinical settings and fail to capture local legal and institutional knowledge. We present Meerkat2, a family of open medical reasoning models at 14B, 32B, and 30B-A3B scales for Korean and English clinical contexts. Meerkat2 is trained on an 837K-example mixture that separates reasoning transfer from high-fidelity local knowledge: clinical reasoning is distilled from strong teachers, while Korean medical statutes are converted into source-grounded question-answer pairs. We also introduce Verbosity-Decoupled Distillation (VDD), which preserves the benefits of long teacher rationales without making verbose generation to the student’s default behavior. Across MedQA and KorMedMCQA, Meerkat2 substantially improves over comparable base models, with the 30B-A3B variant reaching Korean licensing accuracy that matches Qwen3-235B-A22B while using only 3B active parameters. Ablations show that statute-grounded synthesis outperforms standard distillation, and mechanistic analysis reveals that VDD cancels gradients toward verbose generation. These results provide a practical recipe for localized and efficient medical LLMs.

The increasing utilization of patient portals has amplified clinicians’ workloads, primarily due to the necessity of addressing detailed patient inquiries related to their health concerns. The ArchEHR-QA 2025 shared task aims to alleviate this burden by automatically generating accurate, evidence-grounded responses to patients’ questions based on their Electronic Health Records (EHRs). This paper presents a six-stage multi-agent framework specifically developed to identify essential clinical sentences for answering patient questions, leveraging large language models (LLMs). Our approach begins with OpenAI’s o3 model generating focused medical context to guide downstream reasoning. In the subsequent stages, GPT-4.1-based agents assess the relevance of individual sentences, recruit domain experts, and consolidate their judgments to identify essential information for constructing coherent, evidence-grounded responses. Our framework achieved an Overall Factuality score of 62.0 and an Overall Relevance Score of 52.9 on the development set, and corresponding scores of 58.6 and 48.8, respectively, on the test set.