arXiv subjects

Xiao Hu

Publications and source records attributed to Xiao Hu.

At least 109 records · Page 6Linked to original sources

arXiv2025

Large Language Models for Controllable Multi-property Multi-objective Molecule Optimization

In real-world drug design, molecule optimization requires selectively improving multiple molecular properties up to pharmaceutically relevant levels, while maintaining others that already meet such criteria. However, existing computational approaches and instruction-tuned LLMs fail to capture such nuanced property-specific objectives, limiting their practical applicability. To address this, we introduce C-MuMOInstruct, the first instruction-tuning dataset focused on multi-property optimization with explicit, property-specific objectives. Leveraging C-MuMOInstruct, we develop GeLLMO-Cs, a series of instruction-tuned LLMs that can perform targeted property-specific optimization. Our experiments across 5 in-distribution and 5 out-of-distribution tasks show that GeLLMO-Cs consistently outperform strong baselines, achieving up to 126% higher success rate. Notably, GeLLMO-Cs exhibit impressive 0-shot generalization to novel optimization tasks and unseen instructions. This offers a step toward a foundational LLM to support realistic, diverse optimizations with property-specific objectives. C-MuMOInstruct and code are accessible through https://github.com/ninglab/GeLLMO-C.

cs.LG↗

arXiv2025

GeLLMO: Generalizing Large Language Models for Multi-property Molecule Optimization

Despite recent advancements, most computational methods for molecule optimization are constrained to single- or double-property optimization tasks and suffer from poor scalability and generalizability to novel optimization tasks. Meanwhile, Large Language Models (LLMs) demonstrate remarkable out-of-domain generalizability to novel tasks. To demonstrate LLMs' potential for molecule optimization, we introduce MuMOInstruct, the first high-quality instruction-tuning dataset specifically focused on complex multi-property molecule optimization tasks. Leveraging MuMOInstruct, we develop GeLLMOs, a series of instruction-tuned LLMs for molecule optimization. Extensive evaluations across 5 in-domain and 5 out-of-domain tasks demonstrate that GeLLMOs consistently outperform state-of-the-art baselines. GeLLMOs also exhibit outstanding zero-shot generalization to unseen tasks, significantly outperforming powerful closed-source LLMs. Such strong generalizability demonstrates the tremendous potential of GeLLMOs as foundational models for molecule optimization, thereby tackling novel optimization tasks without resource-intensive retraining. MuMOInstruct, models, and code are accessible through https://github.com/ninglab/GeLLMO.

cs.LG↗

arXiv2025

Why Distillation can Outperform Zero-RL: The Role of Flexible Reasoning

Reinforcement learning (RL) has played an important role in improving the reasoning ability of large language models (LLMs). Some studies apply RL directly to \textit{smaller} base models (known as zero-RL) and also achieve notable progress. However, in this paper, we show that using only 920 examples, a simple distillation method based on the base model can clearly outperform zero-RL, which typically requires much more data and computational cost. By analyzing the token frequency in model outputs, we find that the distilled model shows more flexible reasoning. It uses anthropomorphic tokens and logical connectors much more often than the zero-RL model. Further analysis reveals that distillation enhances the presence of two advanced cognitive behaviors: Multi-Perspective Thinking or Attempting and Metacognitive Awareness. Frequent occurrences of these two advanced cognitive behaviors give rise to flexible reasoning, which is essential for solving complex reasoning problems, while zero-RL fails to significantly boost the frequency of these behaviors.

cs.AI↗

arXiv2025

Tactile-based Reinforcement Learning for Adaptive Grasping under Observation Uncertainties

Robotic manipulation in industrial scenarios such as construction commonly faces uncertain observations in which the state of the manipulating object may not be accurately captured due to occlusions and partial observables. For example, object status estimation during pipe assembly, rebar installation, and electrical installation can be impacted by observation errors. Traditional vision-based grasping methods often struggle to ensure robust stability and adaptability. To address this challenge, this paper proposes a tactile simulator that enables a tactile-based adaptive grasping method to enhance grasping robustness. This approach leverages tactile feedback combined with the Proximal Policy Optimization (PPO) reinforcement learning algorithm to dynamically adjust the grasping posture, allowing adaptation to varying grasping conditions under inaccurate object state estimations. Simulation results demonstrate that the proposed method effectively adapts grasping postures, thereby improving the success rate and stability of grasping tasks.

cs.RO↗

arXiv2025

Offline Guarded Safe Reinforcement Learning for Medical Treatment Optimization Strategies

When applying offline reinforcement learning (RL) in healthcare scenarios, the out-of-distribution (OOD) issues pose significant risks, as inappropriate generalization beyond clinical expertise can result in potentially harmful recommendations. While existing methods like conservative Q-learning (CQL) attempt to address the OOD issue, their effectiveness is limited by only constraining action selection by suppressing uncertain actions. This action-only regularization imitates clinician actions that prioritize short-term rewards, but it fails to regulate downstream state trajectories, thereby limiting the discovery of improved long-term treatment strategies. To safely improve policy beyond clinician recommendations while ensuring that state-action trajectories remain in-distribution, we propose \textit{Offline Guarded Safe Reinforcement Learning} ($\mathsf{OGSRL}$), a theoretically grounded model-based offline RL framework. $\mathsf{OGSRL}$ introduces a novel dual constraint mechanism for improving policy with reliability and safety. First, the OOD guardian is established to specify clinically validated regions for safe policy exploration. By constraining optimization within these regions, it enables the reliable exploration of treatment strategies that outperform clinician behavior by leveraging the full patient state history, without drifting into unsupported state-action trajectories. Second, we introduce a safety cost constraint that encodes medical knowledge about physiological safety boundaries, providing domain-specific safeguards even in areas where training data might contain potentially unsafe interventions. Notably, we provide theoretical guarantees on safety and near-optimality: policies that satisfy these constraints remain in safe and reliable regions and achieve performance close to the best possible policy supported by the data.

cs.LG↗

arXiv2025

Early Risk Prediction of Pediatric Cardiac Arrest from Electronic Health Records via Multimodal Fused Transformer

Early prediction of pediatric cardiac arrest (CA) is critical for timely intervention in high-risk intensive care settings. We introduce PedCA-FT, a novel transformer-based framework that fuses tabular view of EHR with the derived textual view of EHR to fully unleash the interactions of high-dimensional risk factors and their dynamics. By employing dedicated transformer modules for each modality view, PedCA-FT captures complex temporal and contextual patterns to produce robust CA risk estimates. Evaluated on a curated pediatric cohort from the CHOA-CICU database, our approach outperforms ten other artificial intelligence models across five key performance metrics and identifies clinically meaningful risk factors. These findings underscore the potential of multimodal fusion techniques to enhance early CA detection and improve patient care.

cs.LG↗

arXiv2025

Dust Concentration Via Coupled Vertical Settling and Radial Migration in Substructured Non-Ideal MHD Discs and Early Planet Formation

We investigate the dynamics of dust concentration in actively accreting, substructured, non-ideal MHD wind-launching disks using 2D and 3D simulations incorporating pressureless dust fluids of various grain sizes and their aerodynamic feedback on gas dynamics. Our results reveal that mm/cm-sized grains are preferentially concentrated within the inner 5-10 au of the disk, where the dust-to-gas surface density ratio (local metalicity Z) significantly exceeds the canonical 0.01, reaching values up to 0.25. This enhancement arises from the interplay of dust settling and complex gas flows in the meridional plane, including midplane accretion streams at early times, midplane expansion driven by magnetically braked surface accretion at later times, and vigorous meridional circulation in spontaneously formed gas rings. The resulting size-dependent dust distribution has a strong spatial variation, with large grains preferentially accumulating in dense rings, particularly in the inner disk, while being depleted in low-density gas gaps. In 3D, these rings and gaps are unstable to Rossby wave instability (RWI), generating arc-shaped vortices that stand out more prominently than their gas counterparts in the inner disk because of preferential dust concentration at small radii. The substantial local enhancement of the dust relative to the gas could promote planetesimal formation via streaming instability, potentially aided by the "azimuthal drift" streaming instability (AdSI) that operates efficiently in accreting disks and a lower Toomre Q expected in younger disks. Our findings suggest that actively accreting young disks may provide favorable conditions for early planetesimal formation, which warrants further investigation.

astro-ph.EP↗

arXiv2025

R1-Reward: Training Multimodal Reward Model Through Stable Reinforcement Learning

Multimodal Reward Models (MRMs) play a crucial role in enhancing the performance of Multimodal Large Language Models (MLLMs). While recent advancements have primarily focused on improving the model structure and training data of MRMs, there has been limited exploration into the effectiveness of long-term reasoning capabilities for reward modeling and how to activate these capabilities in MRMs. In this paper, we explore how Reinforcement Learning (RL) can be used to improve reward modeling. Specifically, we reformulate the reward modeling problem as a rule-based RL task. However, we observe that directly applying existing RL algorithms, such as Reinforce++, to reward modeling often leads to training instability or even collapse due to the inherent limitations of these algorithms. To address this issue, we propose the StableReinforce algorithm, which refines the training loss, advantage estimation strategy, and reward design of existing RL methods. These refinements result in more stable training dynamics and superior performance. To facilitate MRM training, we collect 200K preference data from diverse datasets. Our reward model, R1-Reward, trained using the StableReinforce algorithm on this dataset, significantly improves performance on multimodal reward modeling benchmarks. Compared to previous SOTA models, R1-Reward achieves a $8.4\%$ improvement on the VL Reward-Bench and a $14.3\%$ improvement on the Multimodal Reward Bench. Moreover, with more inference compute, R1-Reward's performance is further enhanced, highlighting the potential of RL algorithms in optimizing MRMs.

cs.CV↗

arXiv2025

Observational Signatures of Disk Winds in Protoplanetary Disks: Differentiating Magnetized and Photoevaporative Outflows With Fully Coupled Thermochemistry

Magnetized winds and photoevaporative winds are critical in shaping protoplanetary disk evolution. Using 2D axisymmetric (magneto-)hydrodynamic simulations with Athena++ implementing fully coupled thermochemistry, we investigate the signatures of the two winds in CO and [C~I] ALMA observations, and examine the potential to distinguish the origins. Our simulations reveal fundamental differences between the two winds: magnetized winds are colder and denser, exhibiting super-Keplerian rotation with small poloidal velocities of $\lesssim 1~{\rm km~s}^{-1}$ in the atmosphere ($z/R\gtrsim0.45$), while photoevaporative winds are hotter and less dense, exhibiting sub-Keplerian rotation with higher poloidal velocity of several ${\rm km~s}^{-1}$. In addition to previously identified factors like thermal pressure gradient and disk's self-gravity, we demonstrate that magnetic tension/pressure and advection significantly influence rotational velocities of the gas in the wind, which lead to emission patterns that are distinct from Keplerian rotation in synthetic ALMA observations. Magnetized winds are visible in CO channel maps when wind loss rates are $\gtrsim10^{-8}~M_\odot~{\rm yr}^{-1}$. When wind loss rates are lower, magnetized winds produce subtle perturbations in channel maps, which resemble the so-called ``velocity kinks'' produced by protoplanets. While photoevaporative winds dissociate CO through strong XUV radiation and thus are weaker in CO, they can create observable ring-like substructures. [C~I] emission is optically thin and could be most effective at detecting both winds in disks with high gas mass and/or high [C~I] abundance. Due to the spatially extended nature of the winds, using a large beam ($\simeq0.4$" for disks in nearby star-forming regions) will be helpful regardless of the tracer used.

astro-ph.EP↗

arXiv2025

Large Language Model Empowered Privacy-Protected Framework for PHI Annotation in Clinical Notes

The de-identification of private information in medical data is a crucial process to mitigate the risk of confidentiality breaches, particularly when patient personal details are not adequately removed before the release of medical records. Although rule-based and learning-based methods have been proposed, they often struggle with limited generalizability and require substantial amounts of annotated data for effective performance. Recent advancements in large language models (LLMs) have shown significant promise in addressing these issues due to their superior language comprehension capabilities. However, LLMs present challenges, including potential privacy risks when using commercial LLM APIs and high computational costs for deploying open-source LLMs locally. In this work, we introduce LPPA, an LLM-empowered Privacy-Protected PHI Annotation framework for clinical notes, targeting the English language. By fine-tuning LLMs locally with synthetic notes, LPPA ensures strong privacy protection and high PHI annotation accuracy. Extensive experiments demonstrate LPPA's effectiveness in accurately de-identifying private information, offering a scalable and efficient solution for enhancing patient privacy protection.

cs.CR↗

arXiv2025

Output-Optimal Algorithms for Join-Aggregate Queries

One of the most celebrated results of computing join-aggregate queries defined over commutative semi-rings is the classic Yannakakis algorithm proposed in 1981. It is known that the runtime of the Yannakakis algorithm is $O(N + \OUT)$ for any free-connex query, where $N$ is the input size of the database and $\OUT$ is the output size of the query result. This is already output-optimal. However, only an upper bound $O(N \cdot \OUT)$ on the runtime is known for the large remaining class of acyclic but non-free-connex queries. Alternatively, one can convert a non-free-connex query into a free-connex one using tree decomposition techniques and then run the Yannakakis algorithm. This approach takes $O\left(N^{\#\fnsubw} + \OUT\right)$ time, where $\#\fnsubw$ is the {\em free-connex sub-modular width} of the input query. But, none of these results is known to be output-optimal. In this paper, we show a matching lower and upper bound $Θ\left(N \cdot \OUT^{1- \frac{1}{\fnfhtw}} + \OUT\right)$ for computing general acyclic join-aggregate queries by {\em semiring algorithms, where $\fnfhtw$ is the free-connex fractional hypertree width} of the query. For example, $\fnfhtw=1$ for free-connex queries, $\fnfhtw =2$ for line queries (a.k.a. chain matrix multiplication), and $\fnfhtw=k$ for star queries (a.k.a. star matrix multiplication) with $k$ relations. While this measure has been defined before, we are the first to use it to characterize the output-optimal complexity of acyclic join-aggregate queries. To our knowledge, this has been the first polynomial improvement over the Yannakakis algorithm in the last 40 years and completely resolves the open question of an output-optimal algorithm for computing acyclic join-aggregate queries.

cs.DB↗

arXiv2025

GPT-PPG: A GPT-based Foundation Model for Photoplethysmography Signals

This study introduces a novel application of a Generative Pre-trained Transformer (GPT) model tailored for photoplethysmography (PPG) signals, serving as a foundation model for various downstream tasks. Adapting the standard GPT architecture to suit the continuous characteristics of PPG signals, our approach demonstrates promising results. Our models are pre-trained on our extensive dataset that contains more than 200 million 30s PPG samples. We explored different supervised fine-tuning techniques to adapt our model to downstream tasks, resulting in performance comparable to or surpassing current state-of-the-art (SOTA) methods in tasks like atrial fibrillation detection. A standout feature of our GPT model is its inherent capability to perform generative tasks such as signal denoising effectively, without the need for further fine-tuning. This success is attributed to the generative nature of the GPT framework.

cs.LG↗

arXiv2025

log-RRIM: Yield Prediction via Local-to-global Reaction Representation Learning and Interaction Modeling

Accurate prediction of chemical reaction yields is crucial for optimizing organic synthesis, potentially reducing time and resources spent on experimentation. With the rise of artificial intelligence (AI), there is growing interest in leveraging AI-based methods to accelerate yield predictions without conducting in vitro experiments. We present log-RRIM, an innovative graph transformer-based framework designed for predicting chemical reaction yields. A key feature of log-RRIM is its integration of a cross-attention mechanism that focuses on the interplay between reagents and reaction centers. This design reflects a fundamental principle in chemical reactions: the crucial role of reagents in influencing bond-breaking and formation processes, which ultimately affect reaction yields. log-RRIM also implements a local-to-global reaction representation learning strategy. This approach initially captures detailed molecule-level information and then models and aggregates intermolecular interactions. Through this hierarchical process, log-RRIM effectively captures how different molecular fragments contribute to and influence the overall reaction yield, regardless of their size variations. log-RRIM shows superior performance in our experiments, especially for medium to high-yielding reactions, proving its reliability as a predictor. The framework's sophisticated modeling of reactant-reagent interactions and precise capture of molecular fragment contributions make it a valuable tool for reaction planning and optimization in chemical synthesis. The data and codes of log-RRIM are accessible through https://github.com/ninglab/Yield_log_RRIM.

q-bio.BM↗

arXiv2025

State-of-the-Art Stroke Lesion Segmentation at 1/1000th of Parameters

Efficient and accurate whole-brain lesion segmentation remains a challenge in medical image analysis. In this work, we revisit MeshNet, a parameter-efficient segmentation model, and introduce a novel multi-scale dilation pattern with an encoder-decoder structure. This innovation enables capturing broad contextual information and fine-grained details without traditional downsampling, upsampling, or skip-connections. Unlike previous approaches processing subvolumes or slices, we operate directly on whole-brain $256^3$ MRI volumes. Evaluations on the Aphasia Recovery Cohort (ARC) dataset demonstrate that MeshNet achieves superior or comparable DICE scores to state-of-the-art architectures such as MedNeXt and U-MAMBA at 1/1000th of parameters. Our results validate MeshNet's strong balance of efficiency and performance, making it particularly suitable for resource-limited environments such as web-based applications and opening new possibilities for the widespread deployment of advanced medical image analysis tools.

eess.IV↗

arXiv2025

Fusion of ECG Foundation Model Embeddings to Improve Early Detection of Acute Coronary Syndromes

Acute Coronary Syndrome (ACS) is a life-threatening cardiovascular condition where early and accurate diagnosis is critical for effective treatment and improved patient outcomes. This study explores the use of ECG foundation models, specifically ST-MEM and ECG-FM, to enhance ACS risk assessment using prehospital ECG data collected in ambulances. Both models leverage self-supervised learning (SSL), with ST-MEM using a reconstruction-based approach and ECG-FM employing contrastive learning, capturing unique spatial and temporal ECG features. We evaluate the performance of these models individually and through a fusion approach, where their embeddings are combined for enhanced prediction. Results demonstrate that both foundation models outperform a baseline ResNet-50 model, with the fusion-based approach achieving the highest performance (AUROC: 0.843 +/- 0.006, AUCPR: 0.674 +/- 0.012). These findings highlight the potential of ECG foundation models for early ACS detection and motivate further exploration of advanced fusion strategies to maximize complementary feature utilization.

eess.SP↗

arXiv2025

Data Center Cooling System Optimization Using Offline Reinforcement Learning

The recent advances in information technology and artificial intelligence have fueled a rapid expansion of the data center (DC) industry worldwide, accompanied by an immense appetite for electricity to power the DCs. In a typical DC, around 30~40% of the energy is spent on the cooling system rather than on computer servers, posing a pressing need for developing new energy-saving optimization technologies for DC cooling systems. However, optimizing such real-world industrial systems faces numerous challenges, including but not limited to a lack of reliable simulation environments, limited historical data, and stringent safety and control robustness requirements. In this work, we present a novel physics-informed offline reinforcement learning (RL) framework for energy efficiency optimization of DC cooling systems. The proposed framework models the complex dynamical patterns and physical dependencies inside a server room using a purposely designed graph neural network architecture that is compliant with the fundamental time-reversal symmetry. Because of its well-behaved and generalizable state-action representations, the model enables sample-efficient and robust latent space offline policy learning using limited real-world operational data. Our framework has been successfully deployed and verified in a large-scale production DC for closed-loop control of its air-cooling units (ACUs). We conducted a total of 2000 hours of short and long-term experiments in the production DC environment. The results show that our method achieves 14~21% energy savings in the DC cooling system, without any violation of the safety or operational constraints. Our results have demonstrated the significant potential of offline RL in solving a broad range of data-limited, safety-critical real-world industrial control problems.

cs.AI↗

arXiv2025

Charge density waves and pinning by lattice anisotropy in 214 cuprates

The detection of static charge density waves (CDWs) in La$_{2-x}$Sr$_x$CuO$_4$ (LSCO) with x $\sim$ 0.12 at relatively high temperatures has raised the question of what lattice feature pins the CDWs. Some recent structural studies have concluded that some form of monoclinic distortion, indicated by the appearance of certain weak Bragg peaks (type M peaks) at otherwise forbidden positions, are responsible for CDW pinning. As a test of this idea, we present neutron diffraction results for a single crystal of La$_{2-x}$Ba$_x$CuO$_4$ (LBCO) with x = 1/8, which is known to undergo two structural transitions on cooling, from high-temperature tetragonal (HTT) to low-temperature orthorhombic (LTO) near 240 K, involving a collective tilt pattern of the corner-sharing CuO$_6$ octahedra, and from LTO to low temperature tetragonal (LTT) near 56 K, involving a new tilt pattern and the appearance of intensity at peaks of type T. We observe both type M and type T peaks in the LTT phase, while the type M peaks (but not type T) are still present in the LTO phase. Given that CDW order is observed only in the LTT phase of LBCO, it is apparent that the in-plane Cu-O bond anisotropy associated with the octahedral tilt pattern is responsible for charge pinning. We point out that evidence for a similar, but weaker, bond anisotropy has been observed previously in LSCO and should be responsible for CDW pinning there. In the case of LBCO, the monoclinic distortion may help to explain previously-reported magneto-optical evidence for gyrotropic order.

cond-mat.supr-con↗

arXiv2025

Continuous Cardiac Arrest Prediction in ICU using PPG Foundation Model

Non-invasive patient monitoring for tracking and predicting adverse acute health events is an emerging area of research. We pursue in-hospital cardiac arrest (IHCA) prediction using only single-channel finger photoplethysmography (PPG) signals. Our proposed two-stage model Feature Extractor-Aggregator Network (FEAN) leverages powerful representations from pre-trained PPG foundation models (PPG-GPT of size up to 1 Billion) stacked with sequential classification models. We propose two FEAN variants ("1H", "FH") which use the latest one-hour and (max) 24-hour history to make decisions respectively. Our study is the first to present IHCA prediction results in ICU patients using only unimodal (continuous PPG signal) waveform deep representations. With our best model, we obtain an average of 0.79 AUROC over 24~h prediction window before CA event onset with our model peaking performance at 0.82 one hour before CA. We also provide a comprehensive analysis of our model through architectural tuning and PaCMAP visualization of patient health trajectory in latent space.

cs.LG↗