Katsuki Fujisawa

Yuepeng Wang, Ken Kawano, Yongqi Zhou et al.

Medical treatment recommendation poses several challenges to reinforcement learning (RL): patient physiology evolves in continuous time, measurements and interventions are performed at irregular intervals, and treatment effects vary substantially across individuals. Existing RL formulations and simulated environments, however, are based on discrete-time MDP or POMDP abstractions with fixed or pre-specified decision intervals. Thus, it remains difficult to evaluate whether RL methods can handle time-interval-dependent disease progression, personalized treatment response, and safety between consecutive measurement points. To address this gap, we introduce MedGym, a benchmark environment for dynamic treatment recommendation. MedGym models longitudinal patient evolution in a continuous-time framework and constructs a configurable medical RL benchmark from clinical data by using Physics-Informed Neural Networks. The resulting benchmark supports both offline and online RL, and enables direct comparison between discrete-time and continuous-time methods under irregular treatment timing and patient-specific dynamics. Besides, MedGym supports evaluation from clinically important perspectives, including personalization, trajectory-level safety, and the performance gap between model-based offline learning and online deployment. By providing a standardized and configurable benchmark for continuous-time dynamic treatment, MedGym aims to facilitate more realistic and informative evaluation of medical RL methods.

Xun Shen, Yuepeng Wang, Akifumi Wachi et al.

Dynamic medical treatment requires deciding treatment intensity and intervention timing, while patient states evolve continuously and adverse events may occur between clinical interactions. Most existing treatment learning methods assume fixed schedules or enforce safety only at discrete decision points. We propose Interaction-Limited Safe Continuous-Time Reinforcement Learning, a framework that jointly optimizes treatment administration and clinical interaction timing under trajectory-level safety constraints. Our key idea is to reformulate the continuous time treatment problem as an option-based semi-Markov decision process, where each option specifies a continuous-time treatment policy and its duration. We develop a safety-tightening mechanism showing that suitably constructed constraints at interaction times guarantee safety over the full continuous-time trajectory with high probability. We further establish finite-sample guarantees for policy learning from logged treatment trajectories and introduce a practical data-driven conservative surrogate. Experiments show that the proposed adaptive interaction-timing mechanism improves both safety and treatment effectiveness over equidistant interaction schemes across different safe policy optimization methods.

Keiichiro Yamamura, Haruki Sato, Nariaki Tateiwa et al.

Deep learning models are vulnerable to adversarial examples, and adversarial attacks used to generate such examples have attracted considerable research interest. Although existing methods based on the steepest descent have achieved high attack success rates, ill-conditioned problems occasionally reduce their performance. To address this limitation, we utilize the conjugate gradient (CG) method, which is effective for this type of problem, and propose a novel attack algorithm inspired by the CG method, named the Auto Conjugate Gradient (ACG) attack. The results of large-scale evaluation experiments conducted on the latest robust models show that, for most models, ACG was able to find more adversarial examples with fewer iterations than the existing SOTA algorithm Auto-PGD (APGD). We investigated the difference in search performance between ACG and APGD in terms of diversification and intensification, and define a measure called Diversity Index (DI) to quantify the degree of diversity. From the analysis of the diversity using this index, we show that the more diverse search of the proposed method remarkably improves its attack success rate.

Yuma Ichikawa, Keiji Kimura, Akihiro Yoshida et al.

Deploying foundation models is increasingly constrained by memory footprint, latency, and hardware costs. Post-training compression can mitigate these bottlenecks by reducing the precision of model parameters without significantly degrading performance; however, its practical implementation remains challenging as practitioners navigate a fragmented landscape of quantization algorithms, precision budgets, data-driven calibration strategies, and hardware-dependent execution regimes. We present OneComp, an open-source compression framework that transforms this expert workflow into a reproducible, resource-adaptive pipeline. Given a model identifier and available hardware, OneComp automatically inspects the model, plans mixed-precision assignments, and executes progressive quantization stages, ranging from layer-wise compression to block-wise refinement and global refinement. A key architectural choice is treating the first quantized checkpoint as a deployable pivot, ensuring that each subsequent stage improves the same model and that quality increases as more compute is invested. By converting state-of-the-art compression research into an extensible, open-source, hardware-aware pipeline, OneComp bridges the gap between algorithmic innovation and production-grade model deployment.

Keiichiro Yamamura, Issa Oe, Hiroki Ishikura et al.

Deep neural networks are vulnerable to adversarial examples, and adversarial attacks that generate adversarial examples have been studied in this context. Existing studies imply that increasing the diversity of model outputs contributes to improving the attack performance. This study focuses on the Auto Conjugate Gradient (ACG) attack, which is inspired by the conjugate gradient method and has a high diversification performance. We hypothesized that increasing the distance between two consecutive search points would enhance the output diversity. To test our hypothesis, we propose Rescaling-ACG (ReACG), which automatically modifies the two components that significantly affect the distance between two consecutive search points, including the search direction and step size. ReACG showed higher attack performance than that of ACG, and is particularly effective for ImageNet models with several classification classes. Experimental results show that the distance between two consecutive search points enhances the output diversity and may help develop new potent attacks. The code is available at \url{https://github.com/yamamura-k/ReACG}

Yoshihiko Fujisawa, Yuma Ichikawa, Yudai Fujimoto et al.

On-device adaptation of large language models commonly keeps a quantized base model frozen while training and deploying a small, task-specific LoRA adapter. In the unmerged adapter-mode setting, however, the adapter is more than a compact storage module; it introduces an additional dense floating-point branch, maintains a trainable state for local updates, and acts as a unit of communication and hot-swapping.We introduce LoRDBA, a LoRA-compatible adapter that replaces both low-rank factors with binary sign carriers while representing magnitudes through lightweight, channel-wise scales, converting the dense adapter branch into two sign-accumulation matrix multiplications interleaved with channel-wise scaling. A finite-sample analysis shows that reconstruction quality is governed by the residual-to-magnitude ratio of the original LoRA factors. In adapter-mode experiments, LoRDBA outperforms low-bit baselines at matched model sizes while matching fp16 LoRA quality in selected regimes. The unmerged adapter incurs at most 8% prefill latency overhead at matched rank r=16 despite an over 10x reduction in adapter footprint, with moderate training memory overhead of approximately 1.6x that of fp16 LoRA.

Yuma Ichikawa, Yoshihiko Fujisawa, Yudai Fujimoto et al.

For extreme low-bit quantization of large language models (LLMs), Double Binary Factorization (DBF) is attractive as it enables efficient inference without sacrificing accuracy. However, the scaling parameters of DBF are too restrictive; after factoring out signs, all rank components share the same magnitude profile, resulting in performance saturation. We propose Multi-envelope DBF (MDBF), which retains a shared pair of 1-bit sign bases but replaces the single envelope with a rank-$l$ envelope. By sharing sign matrices among envelope components, MDBF effectively maintains a binary carrier and utilizes the limited memory budget for magnitude expressiveness. We also introduce a closed-form initialization and an alternating refinement method to optimize MDBF. Across the LLaMA and Qwen families, MDBF enhances perplexity and zero-shot accuracy over previous binary formats at matched bits per weight while preserving the same deployment-friendly inference primitive.

Mingda Liu, Zhenghan Zhu, Ze'an Miao et al.

Model editing has emerged as a practical approach for mitigating factual errors and outdated knowledge in large language models (LLMs). Among existing methods, the Locate-and-Edit (L&E) paradigm is the dominant framework: it locates MLP parameters implicated in expressing a target fact, and then performs a localized update to rewrite that fact. However, long sequences of edits often trigger abrupt model collapse in L&E beyond a critical point. We empirically identify a strong correlation between collapse and explosive growth of edited MLP weight norms, and formally prove that commonly used L&E update rules can induce exponential norm growth across sequential edits in the absence of explicit norm control. To address this issue, we propose Norm-Anchor Scaling NAS, a plug-and-play norm-constrained strategy. Across extensive experiments, NAS delays the collapse point of representative L&E algorithms by more than 4 times and yields a 72.2% average relative gain in editing performance, requiring only a single additional line of code and incurring negligible computational overhead.

Keiichiro Yamamura, Toru Mitsutake, Hiroki Ishikura et al.

Quadratic Unconstrained Binary Optimization (QUBO)-based suppression in object detection is known to have superiority to conventional Non-Maximum Suppression (NMS), especially for crowded scenes where NMS possibly suppresses the (partially-) occluded true positives with low confidence scores. Whereas existing QUBO formulations are less likely to miss occluded objects than NMS, there is room for improvement because existing QUBO formulations naively consider confidence scores and pairwise scores based on spatial overlap between predictions. This study proposes new QUBO formulations that aim to distinguish whether the overlap between predictions is due to the occlusion of objects or due to redundancy in prediction, i.e., multiple predictions for a single object. The proposed QUBO formulation integrates two features into the pairwise score of the existing QUBO formulation: i) the appearance feature calculated by the image similarity metric and ii) the product of confidence scores. These features are derived from the hypothesis that redundant predictions share a similar appearance feature and (partially-) occluded objects have low confidence scores, respectively. The proposed methods demonstrate significant advancement over state-of-the-art QUBO-based suppression without a notable increase in runtime, achieving up to 4.54 points improvement in mAP and 9.89 points gain in mAR.

Nozomi Hata, Shizuo Kaji, Akihiro Yoshida et al.

Studies on acquiring appropriate continuous representations of discrete objects, such as graphs and knowledge base data, have been conducted by many researchers in the field of machine learning. In this study, we introduce Nested SubSpace (NSS) arrangement, a comprehensive framework for representation learning. We show that existing embedding techniques can be regarded as special cases of the NSS arrangement. Based on the concept of the NSS arrangement, we implement a Disk-ANChor ARrangement (DANCAR), a representation learning method specialized to reproducing general graphs. Numerical experiments have shown that DANCAR has successfully embedded WordNet in ${\mathbb R}^{20}$ with an F1 score of 0.993 in the reconstruction task. DANCAR is also suitable for visualization in understanding the characteristics of graphs.