Choonghan Kim

Choonghan Kim, Hyunmin Hwang, Hangeol Chang et al.

While Large Language Models (LLMs) have shown strong performance on clinical text understanding, they struggle with longitudinal prediction tasks such as dementia prognosis, which require reasoning over complex, non-monotonic symptom trajectories across multiple visits. Standard supervised training lacks explicit annotations for symptom evolution, while direct Reinforcement Learning (RL) is hindered by sparse binary rewards. To address this challenge, we introduce Dementia-R1, an RL-based framework for longitudinal dementia prognosis from unstructured clinical notes. Our approach adopts a Cold-Start RL strategy that pre-trains the model to predict verifiable clinical indices extracted from patient histories, enhancing the capability to reason about disease progression before determining the final clinical status. Extensive experiments show that Dementia-R1 achieves the best overall performance on the AMC real-world unstructured cohort, reaching an AUROC of 84.02% and outperforming models up to 10x larger. The framework also generalizes to Parkinson's disease dementia prediction in an independent hospital cohort, achieving an AUROC of 78.37%. On the ADNI benchmark, our 7B model attains the highest AUROC among all LLM baselines at 83.17%, demonstrating strong longitudinal reasoning over fluctuating cognitive trajectories. Code is available at https://anonymous.4open.science/r/dementiar1-CDB5.

Choonghan Kim, Gary Geunbae Lee

Data-to-text (D2T) generation is the task of generating texts from structured inputs. We observed that when the same target sentence was repeated twice, Transformer (T5) based model generates an output made up of asymmetric sentences from structured inputs. In other words, these sentences were different in length and quality. We call this phenomenon "Asymmetric Generation" and we exploit this in D2T generation. Once asymmetric sentences are generated, we add the first part of the output with a no-repeated-target. As this goes through progressive edit (ProEdit), the recall increases. Hence, this method better covers structured inputs than before editing. ProEdit is a simple but effective way to improve performance in D2T generation and it achieves the new stateof-the-art result on the ToTTo dataset

Hyunmin Hwang, Jaemin Kim, Choonghan Kim et al.

Multi-agent reasoning has shown promise for improving the problem-solving ability of large language models by allowing multiple agents to explore diverse reasoning paths. However, most existing multi-agent methods rely on inference-time debate or aggregation, which can be vulnerable to incorrect peer influence and biased consensus. Moreover, the agents themselves remain static, as their underlying reasoning skills do not evolve across tasks. In this paper, we introduce AgentPSO, a particle-swarm-inspired framework for evolving multi-agent reasoning skills. AgentPSO treats each agent as a particle-like reasoner whose state is a natural-language skill and whose velocity is a semantic update direction, iteratively moving agents toward stronger skill states to improve both individual and collective reasoning performance. Across training iterations, each agent updates its skill by combining its previous velocity, personal-best skill, global-best skill, and a self-reflective direction derived from peer reasoning trajectories. This enables agents to learn reusable reasoning behaviors from both their own experiences and the strongest skills discovered by the population, without updating the parameters of the backbone language model. Experiments on mathematical and general reasoning benchmarks show that AgentPSO improves over static single-agent skills and test-time-only multi-agent reasoning baselines. The evolved skills further transfer across benchmarks and to another backbone model, suggesting that AgentPSO captures reusable reasoning procedures rather than merely optimizing benchmark-specific prompts. Code is open-sourced at https://github.com/HYUNMIN-HWANG/AgentPSO/.

Jaemin Kim, Hangeol Chang, Hyunmin Hwang et al.

Large Language Models (LLMs) have demonstrated remarkable general capabilities, but enhancing skills such as reasoning often demands substantial computational resources and may compromise their generalization. While Parameter-Efficient Fine-Tuning (PEFT) methods offer a more resource-conscious alternative, they typically requires retraining for each LLM backbone due to architectural dependencies. To address these challenges, here we propose Universal Reasoner (UniR) - a single, lightweight, composable, and plug-and-play reasoning module that can be used with any frozen LLM to endow it with specialized reasoning capabilities. Specifically, UniR decomposes the reward into a standalone reasoning module that is trained independently using predefined rewards, effectively translating trajectory-level signals into token-level guidance. Once trained, UniR can be combined with any frozen LLM at inference time by simply adding its output logits to those of the LLM backbone. This additive structure naturally enables modular composition: multiple UniR modules trained for different tasks can be jointly applied by summing their logits, enabling complex reasoning via composition. Experimental results on mathematical reasoning and machine translation tasks show that UniR significantly outperforms existing baseline fine-tuning methods using the Llama3.2 model. Furthermore, UniR demonstrates strong weak-to-strong generalization: reasoning modules trained on smaller models effectively guide much larger LLMs. This makes UniR a cost-efficient, adaptable, and robust solution for enhancing reasoning in LLMs without compromising their core capabilities. Code is open-sourced at https://github.com/hangeol/UniR

Choonghan Kim, Seonhee Cho, Joo Heung Yoon

Background: Large language models (LLMs) are gaining use in clinical settings, but their performance can suffer with incomplete radiology reports. We tested whether multimodal LLMs (using text and images) could improve accuracy and understanding in chest radiography reports, making them more effective for clinical decision support. Purpose: To assess the robustness of LLMs in generating accurate impressions from chest radiography reports using both incomplete data and multimodal data. Material and Methods: We used 300 radiology image-report pairs from the MIMIC-CXR database. Three LLMs (OpenFlamingo, MedFlamingo, IDEFICS) were tested in both text-only and multimodal formats. Impressions were first generated from the full text, then tested by removing 20%, 50%, and 80% of the text. The impact of adding images was evaluated using chest x-rays, and model performance was compared using three metrics with statistical analysis. Results: The text-only models (OpenFlamingo, MedFlamingo, IDEFICS) had similar performance (ROUGE-L: 0.39 vs. 0.21 vs. 0.21; F1RadGraph: 0.34 vs. 0.17 vs. 0.17; F1CheXbert: 0.53 vs. 0.40 vs. 0.40), with OpenFlamingo performing best on complete text (p<0.001). Performance declined with incomplete data across all models. However, adding images significantly boosted the performance of MedFlamingo and IDEFICS (p<0.001), equaling or surpassing OpenFlamingo, even with incomplete text. Conclusion: LLMs may produce low-quality outputs with incomplete radiology data, but multimodal LLMs can improve reliability and support clinical decision-making. Keywords: Large language model; multimodal; semantic analysis; Chest Radiography; Clinical Decision Support;

Seonhee Cho, Choonghan Kim, Jiho Lee et al.

Recent advancements in Large Multimodal Models (LMMs) have attracted interest in their generalization capability with only a few samples in the prompt. This progress is particularly relevant to the medical domain, where the quality and sensitivity of data pose unique challenges for model training and application. However, the dependency on high-quality data for effective in-context learning raises questions about the feasibility of these models when encountering with the inevitable variations and errors inherent in real-world medical data. In this paper, we introduce MID-M, a novel framework that leverages the in-context learning capabilities of a general-domain Large Language Model (LLM) to process multimodal data via image descriptions. MID-M achieves a comparable or superior performance to task-specific fine-tuned LMMs and other general-domain ones, without the extensive domain-specific training or pre-training on multimodal data, with significantly fewer parameters. This highlights the potential of leveraging general-domain LLMs for domain-specific tasks and offers a sustainable and cost-effective alternative to traditional LMM developments. Moreover, the robustness of MID-M against data quality issues demonstrates its practical utility in real-world medical domain applications.