Cong Shen

Cong Shen, Jiawei Luo, Kelin Xia

Geometric deep learning (GDL) has demonstrated huge power and enormous potential in molecular data analysis. However, a great challenge still remains for highly efficient molecular representations. Currently, covalent-bond-based molecular graphs are the de facto standard for representing molecular topology at the atomic level. Here we demonstrate, for the first time, that molecular graphs constructed only from non-covalent bonds can achieve similar or even better results than covalent-bond-based models in molecular property prediction. This demonstrates the great potential of novel molecular representations beyond the de facto standard of covalent-bond-based molecular graphs. Based on the finding, we propose molecular geometric deep learning (Mol-GDL). The essential idea is to incorporate a more general molecular representation into GDL models. In our Mol-GDL, molecular topology is modeled as a series of molecular graphs, each focusing on a different scale of atomic interactions. In this way, both covalent interactions and non-covalent interactions are incorporated into the molecular representation on an equal footing. We systematically test Mol-GDL on fourteen commonly-used benchmark datasets. The results show that our Mol-GDL can achieve a better performance than state-of-the-art (SOTA) methods. Source code and data are available at https://github.com/CS-BIO/Mol-GDL.

Xizixiang Wei, Cong Shen, Jing Yang et al.

We propose a novel communication design, termed random orthogonalization, for federated learning (FL) in a massive multiple-input and multiple-output (MIMO) wireless system. The key novelty of random orthogonalization comes from the tight coupling of FL and two unique characteristics of massive MIMO -- channel hardening and favorable propagation. As a result, random orthogonalization can achieve natural over-the-air model aggregation without requiring transmitter side channel state information (CSI) for the uplink phase of FL, while significantly reducing the channel estimation overhead at the receiver. We extend this principle to the downlink communication phase and develop a simple but highly effective model broadcast method for FL. We also relax the massive MIMO assumption by proposing an enhanced random orthogonalization design for both uplink and downlink FL communications, that does not rely on channel hardening or favorable propagation. Theoretical analyses with respect to both communication and machine learning performance are carried out. In particular, an explicit relationship among the convergence rate, the number of clients, and the number of antennas is established. Experimental results validate the effectiveness and efficiency of random orthogonalization for FL in massive MIMO.

Wei Xiong, Han Zhong, Chengshuai Shi et al.

Offline reinforcement learning (RL) aims at learning an optimal strategy using a pre-collected dataset without further interactions with the environment. While various algorithms have been proposed for offline RL in the previous literature, the minimax optimality has only been (nearly) established for tabular Markov decision processes (MDPs). In this paper, we focus on offline RL with linear function approximation and propose a new pessimism-based algorithm for offline linear MDP. At the core of our algorithm is the uncertainty decomposition via a reference function, which is new in the literature of offline RL under linear function approximation. Theoretical analysis demonstrates that our algorithm can match the performance lower bound up to logarithmic factors. We also extend our techniques to the two-player zero-sum Markov games (MGs), and establish a new performance lower bound for MGs, which tightens the existing result, and verifies the nearly minimax optimality of the proposed algorithm. To the best of our knowledge, these are the first computationally efficient and nearly minimax optimal algorithms for offline single-agent MDPs and MGs with linear function approximation.

Wei Xiong, Han Zhong, Chengshuai Shi et al.

Existing studies on provably efficient algorithms for Markov games (MGs) almost exclusively build on the "optimism in the face of uncertainty" (OFU) principle. This work focuses on a different approach of posterior sampling, which is celebrated in many bandits and reinforcement learning settings but remains under-explored for MGs. Specifically, for episodic two-player zero-sum MGs, a novel posterior sampling algorithm is developed with general function approximation. Theoretical analysis demonstrates that the posterior sampling algorithm admits a $\sqrt{T}$-regret bound for problems with a low multi-agent decoupling coefficient, which is a new complexity measure for MGs, where $T$ denotes the number of episodes. When specialized to linear MGs, the obtained regret bound matches the state-of-the-art results. To the best of our knowledge, this is the first provably efficient posterior sampling algorithm for MGs with frequentist regret guarantees, which enriches the toolbox for MGs and promotes the broad applicability of posterior sampling.

Xizixiang Wei, Tianhao Wang, Ruiquan Huang et al.

We propose a privacy-preserving uplink over-the-air computation (AirComp) method, termed FLORAS, for single-input single-output (SISO) wireless federated learning (FL) systems. From the perspective of communication designs, FLORAS eliminates the requirement of channel state information at the transmitters (CSIT) by leveraging the properties of orthogonal sequences. From the privacy perspective, we prove that FLORAS offers both item-level and client-level differential privacy (DP) guarantees. Moreover, by properly adjusting the system parameters, FLORAS can flexibly achieve different DP levels at no additional cost. A new FL convergence bound is derived which, combined with the privacy guarantees, allows for a smooth tradeoff between the achieved convergence rate and differential privacy levels. Experimental results demonstrate the advantages of FLORAS compared with the baseline AirComp method, and validate that the analytical results can guide the design of privacy-preserving FL with different tradeoff requirements on the model convergence and privacy levels.

Cong Shen, Pingjian Ding, Junjie Wee et al.

Geometric deep learning has demonstrated a great potential in non-Euclidean data analysis. The incorporation of geometric insights into learning architecture is vital to its success. Here we propose a curvature-enhanced graph convolutional network (CGCN) for biomolecular interaction prediction, for the first time. Our CGCN employs Ollivier-Ricci curvature (ORC) to characterize network local structures and to enhance the learning capability of GCNs. More specifically, ORCs are evaluated based on the local topology from node neighborhoods, and further used as weights for the feature aggregation in message-passing procedure. Our CGCN model is extensively validated on fourteen real-world bimolecular interaction networks and a series of simulated data. It has been found that our CGCN can achieve the state-of-the-art results. It outperforms all existing models, as far as we know, in thirteen out of the fourteen real-world datasets and ranks as the second in the rest one. The results from the simulated data show that our CGCN model is superior to the traditional GCN models regardless of the positive-to-negativecurvature ratios, network densities, and network sizes (when larger than 500).

Sihua Wang, Mingzhe Chen, Cong Shen et al.

In this paper, the performance optimization of federated learning (FL), when deployed over a realistic wireless multiple-input multiple-output (MIMO) communication system with digital modulation and over-the-air computation (AirComp) is studied. In particular, a MIMO system is considered in which edge devices transmit their local FL models (trained using their locally collected data) to a parameter server (PS) using beamforming to maximize the number of devices scheduled for transmission. The PS, acting as a central controller, generates a global FL model using the received local FL models and broadcasts it back to all devices. Due to the limited bandwidth in a wireless network, AirComp is adopted to enable efficient wireless data aggregation. However, fading of wireless channels can produce aggregate distortions in an AirComp-based FL scheme. To tackle this challenge, we propose a modified federated averaging (FedAvg) algorithm that combines digital modulation with AirComp to mitigate wireless fading while ensuring the communication efficiency. This is achieved by a joint transmit and receive beamforming design, which is formulated as an optimization problem to dynamically adjust the beamforming matrices based on current FL model parameters so as to minimize the transmitting error and ensure the FL performance. To achieve this goal, we first analytically characterize how the beamforming matrices affect the performance of the FedAvg in different iterations. Based on this relationship, an artificial neural network (ANN) is used to estimate the local FL models of all devices and adjust the beamforming matrices at the PS for future model transmission. The algorithmic advantages and improved performance of the proposed methodologies are demonstrated through extensive numerical experiments.

Shaoxuan Wu, Jingkun Chen, Chong Ma et al.

Computer-aided diagnosis (CAD) has significantly advanced automated chest X-ray diagnosis but remains isolated from clinical workflows and lacks reliable decision support and interpretability. Human-AI collaboration seeks to enhance the reliability of diagnostic models by integrating the behaviors of controllable radiologists. However, the absence of interactive tools seamlessly embedded within diagnostic routines impedes collaboration, while the semantic gap between radiologists' decision-making patterns and model representations further limits clinical adoption. To overcome these limitations, we propose a visual cognition-guided collaborative network (VCC-Net) to achieve the cooperative diagnostic paradigm. VCC-Net centers on visual cognition (VC) and employs clinically compatible interfaces, such as eye-tracking or the mouse, to capture radiologists' visual search traces and attention patterns during diagnosis. VCC-Net employs VC as a spatial cognition guide, learning hierarchical visual search strategies to localize diagnostically key regions. A cognition-graph co-editing module subsequently integrates radiologist VC with model inference to construct a disease-aware graph. The module captures dependencies among anatomical regions and aligns model representations with VC-driven features, mitigating radiologist bias and facilitating complementary, transparent decision-making. Experiments on the public datasets SIIM-ACR, EGD-CXR, and self-constructed TB-Mouse dataset achieved classification accuracies of 88.40%, 85.05%, and 92.41%, respectively. The attention maps produced by VCC-Net exhibit strong concordance with radiologists' gaze distributions, demonstrating a mutual reinforcement of radiologist and model inference. The code is available at https://github.com/IPMI-NWU/VCC-Net.

Wenjing Chen, Ruida Zhou, Chao Tian et al.

We analyze the performance of the Borda counting algorithm in a non-parametric model. The algorithm needs to utilize probabilistic rankings of the items within $m$-sized subsets to accurately determine which items are the overall top-$k$ items in a total of $n$ items. The Borda counting algorithm simply counts the cumulative scores for each item from these partial ranking observations. This generalizes a previous work of a similar nature by Shah et al. using probabilistic pairwise comparison data. The performance of the Borda counting algorithm critically depends on the associated score separation $Δ_k$ between the $k$-th item and the $(k+1)$-th item. Specifically, we show that if $Δ_k$ is greater than certain value, then the top-$k$ items selected by the algorithm is asymptotically accurate almost surely; if $Δ_k$ is below certain value, then the result will be inaccurate with a constant probability. In the special case of $m=2$, i.e., pairwise comparison, the resultant bound is tighter than that given by Shah et al., leading to a reduced gap between the error probability upper and lower bounds. These results are further extended to the approximate top-$k$ selection setting. Numerical experiments demonstrate the effectiveness and accuracy of the Borda counting algorithm, compared with the spectral MLE-based algorithm, particularly when the data does not necessarily follow an assumed parametric model.

Kun Yang, Cong Shen, Jing Yang et al.

The recent development of reinforcement learning (RL) has boosted the adoption of online RL for wireless radio resource management (RRM). However, online RL algorithms require direct interactions with the environment, which may be undesirable given the potential performance loss due to the unavoidable exploration in RL. In this work, we first investigate the use of \emph{offline} RL algorithms in solving the RRM problem. We evaluate several state-of-the-art offline RL algorithms, including behavior constrained Q-learning (BCQ), conservative Q-learning (CQL), and implicit Q-learning (IQL), for a specific RRM problem that aims at maximizing a linear combination {of sum and} 5-percentile rates via user scheduling. We observe that the performance of offline RL for the RRM problem depends critically on the behavior policy used for data collection, and further propose a novel offline RL solution that leverages heterogeneous datasets collected by different behavior policies. We show that with a proper mixture of the datasets, offline RL can produce a near-optimal RL policy even when all involved behavior policies are highly suboptimal.

Jieming Bian, Lei Wang, Kun Yang et al.

Over the past few years, Federated Learning (FL) has become a popular distributed machine learning paradigm. FL involves a group of clients with decentralized data who collaborate to learn a common model under the coordination of a centralized server, with the goal of protecting clients' privacy by ensuring that local datasets never leave the clients and that the server only performs model aggregation. However, in realistic scenarios, the server may be able to collect a small amount of data that approximately mimics the population distribution and has stronger computational ability to perform the learning process. To address this, we focus on the hybrid FL framework in this paper. While previous hybrid FL work has shown that the alternative training of clients and server can increase convergence speed, it has focused on the scenario where clients fully participate and ignores the negative effect of partial participation. In this paper, we provide theoretical analysis of hybrid FL under clients' partial participation to validate that partial participation is the key constraint on convergence speed. We then propose a new algorithm called FedCLG, which investigates the two-fold role of the server in hybrid FL. Firstly, the server needs to process the training steps using its small amount of local datasets. Secondly, the server's calculated gradient needs to guide the participated clients' training and the server's aggregation. We validate our theoretical findings through numerical experiments, which show that our proposed method FedCLG outperforms state-of-the-art methods.

Yujia Mu, Cong Shen

Federated learning (FL) is a popular distributed machine learning (ML) paradigm, but is often limited by significant communication costs and edge device computation capabilities. Federated Split Learning (FSL) preserves the parallel model training principle of FL, with a reduced device computation requirement thanks to splitting the ML model between the server and clients. However, FSL still incurs very high communication overhead due to transmitting the smashed data and gradients between the clients and the server in each global round. Furthermore, the server has to maintain separate models for every client, resulting in a significant computation and storage requirement that grows linearly with the number of clients. This paper tries to solve these two issues by proposing a communication and storage efficient federated and split learning (CSE-FSL) strategy, which utilizes an auxiliary network to locally update the client models while keeping only a single model at the server, hence avoiding the communication of gradients from the server and greatly reducing the server resource requirement. Communication cost is further reduced by only sending the smashed data in selected epochs from the clients. We provide a rigorous theoretical analysis of CSE-FSL that guarantees its convergence for non-convex loss functions. Extensive experimental results demonstrate that CSE-FSL has a significant communication reduction over existing FSL techniques while achieving state-of-the-art convergence and model accuracy, using several real-world FL tasks.

Song Wang, Zihan Chen, Peng Wang et al.

Retrieval-augmented generation (RAG) enhances large language models (LLMs) by integrating external knowledge sources to address their limitations in accessing up-to-date or specialized information. A natural strategy to increase the likelihood of retrieving relevant information is to expand the number of retrieved documents. However, involving more documents could introduce significant noise, as many documents may be irrelevant or misleading, thereby reducing the overall accuracy of the generated responses. To overcome the challenge associated with handling a larger number of documents, we propose WinnowRAG, a novel RAG framework designed to systematically filter out noisy documents while preserving valuable content -- a process we refer to as winnowing. WinnowRAG operates in two stages: In Stage I, we perform query-aware clustering to group similar documents and form distinct topic clusters. Each cluster is assigned to an LLM agent for generating a unique answer. In Stage II, we perform winnowing, wherein a critic LLM evaluates the outputs of multiple agents and iteratively separates useful documents from noisy ones. To retain useful documents when discarding agents, we propose two strategic merging techniques to ensure that only relevant knowledge is used for generating the final response. Crucially, WinnowRAG is model-agnostic and does not require any model fine-tuning, making it easily adaptable to various tasks. Extensive experiments on various realistic datasets demonstrate the effectiveness of WinnowRAG over state-of-the-art baselines.

Chengshuai Shi, Wei Xiong, Cong Shen et al.

Existing theoretical studies on offline reinforcement learning (RL) mostly consider a dataset sampled directly from the target task. In practice, however, data often come from several heterogeneous but related sources. Motivated by this gap, this work aims at rigorously understanding offline RL with multiple datasets that are collected from randomly perturbed versions of the target task instead of from itself. An information-theoretic lower bound is derived, which reveals a necessary requirement on the number of involved sources in addition to that on the number of data samples. Then, a novel HetPEVI algorithm is proposed, which simultaneously considers the sample uncertainties from a finite number of data samples per data source and the source uncertainties due to a finite number of available data sources. Theoretical analyses demonstrate that HetPEVI can solve the target task as long as the data sources collectively provide a good data coverage. Moreover, HetPEVI is demonstrated to be optimal up to a polynomial factor of the horizon length. Finally, the study is extended to offline Markov games and offline robust RL, which demonstrates the generality of the proposed designs and theoretical analyses.

Cong Shen, Xiang Liu, Jiawei Luo et al.

Geometric deep learning (GDL) models have demonstrated a great potential for the analysis of non-Euclidian data. They are developed to incorporate the geometric and topological information of non-Euclidian data into the end-to-end deep learning architectures. Motivated by the recent success of discrete Ricci curvature in graph neural network (GNNs), we propose TorGNN, an analytic Torsion enhanced Graph Neural Network model. The essential idea is to characterize graph local structures with an analytic torsion based weight formula. Mathematically, analytic torsion is a topological invariant that can distinguish spaces which are homotopy equivalent but not homeomorphic. In our TorGNN, for each edge, a corresponding local simplicial complex is identified, then the analytic torsion (for this local simplicial complex) is calculated, and further used as a weight (for this edge) in message-passing process. Our TorGNN model is validated on link prediction tasks from sixteen different types of networks and node classification tasks from three types of networks. It has been found that our TorGNN can achieve superior performance on both tasks, and outperform various state-of-the-art models. This demonstrates that analytic torsion is a highly efficient topological invariant in the characterization of graph structures and can significantly boost the performance of GNNs.

Binchi Zhang, Zihan Chen, Cong Shen et al.

As privacy concerns escalate in the realm of machine learning, data owners now have the option to utilize machine unlearning to remove their data from machine learning models, following recent legislation. To enhance transparency in machine unlearning and avoid potential dishonesty by model providers, various verification strategies have been proposed. These strategies enable data owners to ascertain whether their target data has been effectively unlearned from the model. However, our understanding of the safety issues of machine unlearning verification remains nascent. In this paper, we explore the novel research question of whether model providers can circumvent verification strategies while retaining the information of data supposedly unlearned. Our investigation leads to a pessimistic answer: \textit{the verification of machine unlearning is fragile}. Specifically, we categorize the current verification strategies regarding potential dishonesty among model providers into two types. Subsequently, we introduce two novel adversarial unlearning processes capable of circumventing both types. We validate the efficacy of our methods through theoretical analysis and empirical experiments using real-world datasets. This study highlights the vulnerabilities and limitations in machine unlearning verification, paving the way for further research into the safety of machine unlearning.

Jieming Bian, Cong Shen, Jie Xu

Federated Learning (FL) is a communication-efficient and privacy-preserving distributed machine learning framework that has gained a significant amount of research attention recently. Despite the different forms of FL algorithms (e.g., synchronous FL, asynchronous FL) and the underlying optimization methods, nearly all existing works implicitly assumed the existence of a communication infrastructure that facilitates the direct communication between the server and the clients for the model data exchange. This assumption, however, does not hold in many real-world applications that can benefit from distributed learning but lack a proper communication infrastructure (e.g., smart sensing in remote areas). In this paper, we propose a novel FL framework, named FedEx (short for FL via Model Express Delivery), that utilizes mobile transporters (e.g., Unmanned Aerial Vehicles) to establish indirect communication channels between the server and the clients. Two algorithms, called FedEx-Sync and FedEx-Async, are developed depending on whether the transporters adopt a synchronized or an asynchronized schedule. Even though the indirect communications introduce heterogeneous delays to clients for both the global model dissemination and the local model collection, we prove the convergence of both versions of FedEx. The convergence analysis subsequently sheds lights on how to assign clients to different transporters and design the routes among the clients. The performance of FedEx is evaluated through experiments in a simulated network on two public datasets.

Jieming Bian, Cong Shen, Jie Xu

Federated Learning (FL) is a machine learning approach that enables the creation of shared models for powerful applications while allowing data to remain on devices. This approach provides benefits such as improved data privacy, security, and reduced latency. However, in some systems, direct communication between clients and servers may not be possible, such as remote areas without proper communication infrastructure. To overcome this challenge, a new framework called FedEx (Federated Learning via Model Express Delivery) is proposed. This framework employs mobile transporters, such as UAVs, to establish indirect communication channels between the server and clients. These transporters act as intermediaries and allow for model information exchange. The use of indirect communication presents new challenges for convergence analysis and optimization, as the delay introduced by the transporters' movement creates issues for both global model dissemination and local model collection. To address this, two algorithms, FedEx-Sync and FedEx-Async, are proposed for synchronized and asynchronized learning at the transporter level. Additionally, a bi-level optimization algorithm is proposed to solve the joint client assignment and route planning problem. Experimental validation using two public datasets in a simulated network demonstrates consistent results with the theory, proving the efficacy of FedEx.

Donghao Li, Ruiquan Huang, Cong Shen et al.

This paper investigates conservative exploration in reinforcement learning where the performance of the learning agent is guaranteed to be above a certain threshold throughout the learning process. It focuses on the tabular episodic Markov Decision Process (MDP) setting that has finite states and actions. With the knowledge of an existing safe baseline policy, an algorithm termed as StepMix is proposed to balance the exploitation and exploration while ensuring that the conservative constraint is never violated in each episode with high probability. StepMix features a unique design of a mixture policy that adaptively and smoothly interpolates between the baseline policy and the optimistic policy. Theoretical analysis shows that StepMix achieves near-optimal regret order as in the constraint-free setting, indicating that obeying the stringent episode-wise conservative constraint does not compromise the learning performance. Besides, a randomization-based EpsMix algorithm is also proposed and shown to achieve the same performance as StepMix. The algorithm design and theoretical analysis are further extended to the setting where the baseline policy is not given a priori but must be learned from an offline dataset, and it is proved that similar conservative guarantee and regret can be achieved if the offline dataset is sufficiently large. Experiment results corroborate the theoretical analysis and demonstrate the effectiveness of the proposed conservative exploration strategies.

Li Fan, Ruida Zhou, Chao Tian et al.

We study a federated linear bandits model, where $M$ clients communicate with a central server to solve a linear contextual bandits problem with finite adversarial action sets that may be different across clients. To address the unique challenges of adversarial finite action sets, we propose the FedSupLinUCB algorithm, which extends the principles of SupLinUCB and OFUL algorithms in linear contextual bandits. We prove that FedSupLinUCB achieves a total regret of $\tilde{O}(\sqrt{d T})$, where $T$ is the total number of arm pulls from all clients, and $d$ is the ambient dimension of the linear model. This matches the minimax lower bound and thus is order-optimal (up to polylog terms). We study both asynchronous and synchronous cases and show that the communication cost can be controlled as $O(d M^2 \log(d)\log(T))$ and $O(\sqrt{d^3 M^3} \log(d))$, respectively. The FedSupLinUCB design is further extended to two scenarios: (1) variance-adaptive, where a total regret of $\tilde{O} (\sqrt{d \sum \nolimits_{t=1}^{T} σ_t^2})$ can be achieved with $σ_t^2$ being the noise variance of round $t$; and (2) adversarial corruption, where a total regret of $\tilde{O}(\sqrt{dT} + d C_p)$ can be achieved with $C_p$ being the total corruption budget. Experiment results corroborate the theoretical analysis and demonstrate the effectiveness of FedSupLinUCB on both synthetic and real-world datasets.

Zhoubin Kou, Zihan Chen, Jing Yang et al.

Split Federated Learning (SFL) enables collaborative training between resource-constrained edge devices and a compute-rich server. Communication overhead is a central issue in SFL and can be mitigated with auxiliary networks. Yet, the fundamental client-side computation challenge remains, as back-propagation requires substantial memory and computation costs, severely limiting the scale of models that edge devices can support. To enable more resource-efficient client computation and reduce the client-server communication, we propose HERON-SFL, a novel hybrid optimization framework that integrates zeroth-order (ZO) optimization for local client training while retaining first-order (FO) optimization on the server. With the assistance of auxiliary networks, ZO updates enable clients to approximate local gradients using perturbed forward-only evaluations per step, eliminating memory-intensive activation caching and avoiding explicit gradient computation in the traditional training process. Leveraging the low effective rank assumption, we theoretically prove that HERON-SFL's convergence rate is independent of model dimensionality, addressing a key scalability concern common to ZO algorithms. Empirically, on ResNet training and language model (LM) fine-tuning tasks, HERON-SFL matches benchmark accuracy while reducing client peak memory by up to 64% and client-side compute cost by up to 33% per step, substantially expanding the range of models that can be trained or adapted on resource-limited devices.

Donghao Li, Chengshuai Shi, Weijuan Ou et al.

Prompt engineering has become central to eliciting the capabilities of large language models (LLMs). At its core lies prompt selection -- efficiently identifying the most effective prompts. However, most prior investigations overlook a key challenge: the inherently multi-faceted nature of prompt performance, which cannot be captured by a single metric. To fill this gap, we study the multi-objective prompt selection problem under two practical settings: Pareto prompt set recovery and best feasible prompt identification. Casting the problem into the pure-exploration bandits framework, we adapt provably efficient algorithms from multi-objective bandits and further introduce a novel design for best feasible arm identification in structured bandits, with theoretical guarantees on the identification error in the linear case. Extensive experiments across multiple LLMs show that the bandit-based approaches yield significant improvements over baselines, establishing a principled and efficient framework for multi-objective prompt optimization.

Boyuan Li, Zihao Peng, Yafei Li et al.

Federated Learning (FL) can be coordinated under the orchestration of a central server to collaboratively build a privacy-preserving model without the need for data exchange. However, participant data heterogeneity leads to local optima divergence, subsequently affecting convergence outcomes. Recent research has focused on global sharpness-aware minimization (SAM) and dynamic regularization techniques to enhance consistency between global and local generalization and optimization objectives. Nonetheless, the estimation of global SAM introduces additional computational and memory overhead, while dynamic regularization suffers from bias in the local and global dual variables due to training isolation. In this paper, we propose a novel FL algorithm, FedTOGA, designed to consider optimization and generalization objectives while maintaining minimal uplink communication overhead. By linking local perturbations to global updates, global generalization consistency is improved. Additionally, global updates are used to correct local dynamic regularizers, reducing dual variables bias and enhancing optimization consistency. Global updates are passively received by clients, reducing overhead. We also propose neighborhood perturbation to approximate local perturbation, analyzing its strengths and limitations. Theoretical analysis shows FedTOGA achieves faster convergence $O(1/T)$ under non-convex functions. Empirical studies demonstrate that FedTOGA outperforms state-of-the-art algorithms, with a 1\% accuracy increase and 30\% faster convergence, achieving state-of-the-art.

Songwei Dong, Zihan Chen, Chengshuai Shi et al.

Memory plays a central role in enabling large language models (LLMs) to operate over sequential tasks by accumulating and reusing experience over time. However, existing evaluations of LLM memory mostly rely on aggregate metrics such as final hold-out accuracy or cumulative online performance, which can obscure critical failure modes such as forgetting and negative transfer. In this paper, we introduce SeqMem-Eval, a diagnostic evaluation framework for sequentially evolving LLM memory. Drawing inspiration from continual learning, it targets a test-time setting in which memory is external, prompt-mediated, and updated without modifying model parameters. Rather than focusing only on final performance, SeqMem-Eval evaluates how memory states evolve, generalize, consolidate experience, and retain useful information during sequential inference. Specifically, it measures online utility, hold-out generalization, backward transfer, and forgetting, providing a finer-grained view of memory quality. Through extensive experiments across diverse tasks and memory methods, we show that higher final or cumulative accuracy does not necessarily imply better memory quality: many methods exhibit strong performance gains while suffering from substantial forgetting or negative transfer. Moreover, different memory designs exhibit distinct trade-offs between adaptability and stability that remain invisible under standard evaluation metrics.

Wei Shen, Jiawei Zhang, Minhui Huang et al.

We study bilevel optimization problems where the lower-level problems are strongly convex and have coupled linear constraints. To overcome the potential non-smoothness of the hyper-objective and the computational challenges associated with the Hessian matrix, we utilize penalty and augmented Lagrangian methods to reformulate the original problem as a single-level one. Especially, we establish a strong theoretical connection between the reformulated function and the original hyper-objective by characterizing the closeness of their values and derivatives. Based on this reformulation, we propose a single-loop, first-order algorithm for linearly constrained bilevel optimization (SFLCB). We provide rigorous analyses of its non-asymptotic convergence rates, showing an improvement over prior double-loop algorithms -- form $O(ε^{-3}\log(ε^{-1}))$ to $O(ε^{-3})$. The experiments corroborate our theoretical findings and demonstrate the practical efficiency of the proposed SFLCB algorithm. Simulation code is provided at https://github.com/ShenGroup/SFLCB.

Wei Shen, Minhui Huang, Jiawei Zhang et al.

In recent years, federated minimax optimization has attracted growing interest due to its extensive applications in various machine learning tasks. While Smoothed Alternative Gradient Descent Ascent (Smoothed-AGDA) has proved its success in centralized nonconvex minimax optimization, how and whether smoothing technique could be helpful in federated setting remains unexplored. In this paper, we propose a new algorithm termed Federated Stochastic Smoothed Gradient Descent Ascent (FESS-GDA), which utilizes the smoothing technique for federated minimax optimization. We prove that FESS-GDA can be uniformly used to solve several classes of federated minimax problems and prove new or better analytical convergence results for these settings. We showcase the practical efficiency of FESS-GDA in practical federated learning tasks of training generative adversarial networks (GANs) and fair classification.

Di Wu, Chengshuai Shi, Jing Yang et al.

Reinforcement Learning from Human Feedback (RLHF) has become a cornerstone technique for post-training large language models. While most existing approaches rely on the reverse KL-regularization, recent empirical studies have begun exploring alternative divergences (e.g., forward KL, chi-squared) as regularizers in RLHF. However, a unified theoretical understanding of general $f$-divergence regularization remains under-explored. To fill this gap, this work develops a comprehensive theoretical framework for online RLHF with a general $f$-divergence regularized objective. Rather than treating each possible divergence function individually, we adopt a holistic perspective across the entire function class and propose two algorithms based on distinct sampling principles. The first extends the classical optimism principle with a carefully designed exploration bonus, while the second introduces a new method that exploits the sensitivity of the optimal policy to reward perturbations under $f$-divergence regularization. Theoretical analysis shows that $O(\log T)$ regret and $O(1/T)$ sub-optimality gap are achievable, establishing provable efficiency of both algorithms and, to the best of our knowledge, the first performance bounds for online RLHF under general $f$-divergence regularization.

Chengshuai Shi, Kun Yang, Zihan Chen et al.

The remarkable instruction-following capability of large language models (LLMs) has sparked a growing interest in automatically finding good prompts, i.e., prompt optimization. Most existing works follow the scheme of selecting from a pre-generated pool of candidate prompts. However, these designs mainly focus on the generation strategy, while limited attention has been paid to the selection method. Especially, the cost incurred during the selection (e.g., accessing LLM and evaluating the responses) is rarely explicitly considered. To overcome this limitation, this work provides a principled framework, TRIPLE, to efficiently perform prompt selection under an explicit budget constraint. TRIPLE is built on a novel connection established between prompt optimization and fixed-budget best arm identification (BAI-FB) in multi-armed bandits (MAB); thus, it is capable of leveraging the rich toolbox from BAI-FB systematically and also incorporating unique characteristics of prompt optimization. Extensive experiments on multiple well-adopted tasks using various LLMs demonstrate the remarkable performance improvement of TRIPLE over baselines while satisfying the limited budget constraints. As an extension, variants of TRIPLE are proposed to efficiently select examples for few-shot prompts, also achieving superior empirical performance.

Pranshav Gajjar, Cong Shen, Vijay K Shah

This paper introduces Tele-LLM-Hub, a user friendly low-code solution for rapid prototyping and deployment of context aware multi-agent (MA) Large Language Model (LLM) systems tailored for 5G and beyond. As telecom wireless networks become increasingly complex, intelligent LLM applications must share a domainspecific understanding of network state. We propose TeleMCP, the Telecom Model Context Protocol, to enable structured and context-rich communication between agents in telecom environments. Tele-LLM-Hub actualizes TeleMCP through a low-code interface that supports agent creation, workflow composition, and interaction with software stacks such as srsRAN. Key components include a direct chat interface, a repository of pre-built systems, an Agent Maker leveraging finetuning with our RANSTRUCT framework, and an MA-Maker for composing MA workflows. The goal of Tele-LLM-Hub is to democratize the design of contextaware MA systems and accelerate innovation in next-generation wireless networks.

Song Wang, Yushun Dong, Binchi Zhang et al.

Graph Machine Learning (Graph ML) has witnessed substantial advancements in recent years. With their remarkable ability to process graph-structured data, Graph ML techniques have been extensively utilized across diverse applications, including critical domains like finance, healthcare, and transportation. Despite their societal benefits, recent research highlights significant safety concerns associated with the widespread use of Graph ML models. Lacking safety-focused designs, these models can produce unreliable predictions, demonstrate poor generalizability, and compromise data confidentiality. In high-stakes scenarios such as financial fraud detection, these vulnerabilities could jeopardize both individuals and society at large. Therefore, it is imperative to prioritize the development of safety-oriented Graph ML models to mitigate these risks and enhance public confidence in their applications. In this survey paper, we explore three critical aspects vital for enhancing safety in Graph ML: reliability, generalizability, and confidentiality. We categorize and analyze threats to each aspect under three headings: model threats, data threats, and attack threats. This novel taxonomy guides our review of effective strategies to protect against these threats. Our systematic review lays a groundwork for future research aimed at developing practical, safety-centered Graph ML models. Furthermore, we highlight the significance of safe Graph ML practices and suggest promising avenues for further investigation in this crucial area.

Zihan Chen, Song Wang, Zhen Tan et al.

The rapid advancements in large Language models (LLMs) have significantly enhanced their reasoning capabilities, driven by various strategies such as multi-agent collaboration. However, unlike the well-established performance improvements achieved through scaling data and model size, the scaling of reasoning in LLMs is more complex and can even negatively impact reasoning performance, introducing new challenges in model alignment and robustness. In this survey, we provide a comprehensive examination of scaling in LLM reasoning, categorizing it into multiple dimensions and analyzing how and to what extent different scaling strategies contribute to improving reasoning capabilities. We begin by exploring scaling in input size, which enables LLMs to process and utilize more extensive context for improved reasoning. Next, we analyze scaling in reasoning steps that improves multi-step inference and logical consistency. We then examine scaling in reasoning rounds, where iterative interactions refine reasoning outcomes. Furthermore, we discuss scaling in training-enabled reasoning, focusing on optimization through iterative model improvement. Finally, we review applications of scaling across domains and outline future directions for further advancing LLM reasoning. By synthesizing these diverse perspectives, this survey aims to provide insights into how scaling strategies fundamentally enhance the reasoning capabilities of LLMs and further guide the development of next-generation AI systems.

Zihan Chen, Song Wang, Zhen Tan et al.

In-Context Learning (ICL) empowers Large Language Models (LLMs) to tackle diverse tasks by incorporating multiple input-output examples, known as demonstrations, into the input of LLMs. More recently, advancements in the expanded context windows of LLMs have led to many-shot ICL, which uses hundreds of demonstrations and outperforms few-shot ICL, which relies on fewer examples. However, this approach is often hindered by the high cost of obtaining large amounts of labeled data. To address this challenge, we propose Many-Shot Adaptive Pseudo-LabEling, namely MAPLE, a novel influence-based many-shot ICL framework that utilizes pseudo-labeled samples to compensate for the lack of label information. We first identify a subset of impactful unlabeled samples and perform pseudo-labeling on them by querying LLMs. These pseudo-labeled samples are then adaptively selected and tailored to each test query as input to improve the performance of many-shot ICL, without significant labeling costs. Extensive experiments on real-world datasets demonstrate the effectiveness of our framework, showcasing its ability to enhance LLM adaptability and performance with limited labeled data.

Kun Yang, Shu-ping Yeh, Menglei Zhang et al.

Dynamic radio resource management (RRM) in wireless networks presents significant challenges, particularly in the context of Radio Access Network (RAN) slicing. This technology, crucial for catering to varying user requirements, often grapples with complex optimization scenarios. Existing Reinforcement Learning (RL) approaches, while achieving good performance in RAN slicing, typically rely on online algorithms or behavior cloning. These methods necessitate either continuous environmental interactions or access to high-quality datasets, hindering their practical deployment. Towards addressing these limitations, this paper introduces offline RL to solving the RAN slicing problem, marking a significant shift towards more feasible and adaptive RRM methods. We demonstrate how offline RL can effectively learn near-optimal policies from sub-optimal datasets, a notable advancement over existing practices. Our research highlights the inherent flexibility of offline RL, showcasing its ability to adjust policy criteria without the need for additional environmental interactions. Furthermore, we present empirical evidence of the efficacy of offline RL in adapting to various service-level requirements, illustrating its potential in diverse RAN slicing scenarios.

Wei Shen, Ruida Zhou, Jing Yang et al.

Although transformers have demonstrated impressive capabilities for in-context learning (ICL) in practice, theoretical understanding of the underlying mechanism that allows transformers to perform ICL is still in its infancy. This work aims to theoretically study the training dynamics of transformers for in-context classification tasks. We demonstrate that, for in-context classification of Gaussian mixtures under certain assumptions, a single-layer transformer trained via gradient descent converges to a globally optimal model at a linear rate. We further quantify the impact of the training and testing prompt lengths on the ICL inference error of the trained transformer. We show that when the lengths of training and testing prompts are sufficiently large, the prediction of the trained transformer approaches the ground truth distribution of the labels. Experimental results corroborate the theoretical findings.

Zihan Chen, Bike Xie, Jundong Li et al.

Large Language Models (LLMs) have demonstrated remarkable success across a wide range of language tasks, but their deployment on edge devices remains challenging due to the substantial memory requirements imposed by their large parameter sizes. Weight-only quantization presents a promising solution to reduce the memory footprint of LLMs. However, existing approaches primarily focus on integer-bit quantization, limiting their adaptability to fractional-bit quantization tasks and preventing the full utilization of available storage space on devices. In this paper, we introduce Channel-Wise Mixed-Precision Quantization (CMPQ), a novel mixed-precision quantization method that allocates quantization precision in a channel-wise pattern based on activation distributions. By assigning different precision levels to different weight channels, CMPQ can adapt to any bit-width constraint. CMPQ employs a non-uniform quantization strategy and incorporates two outlier extraction techniques that collaboratively preserve the critical information, thereby minimizing the quantization loss. Experiments on different sizes of LLMs demonstrate that CMPQ not only enhances performance in integer-bit quantization tasks but also achieves significant performance gains with a modest increase in memory usage. CMPQ thus represents an adaptive and effective approach to LLM quantization, offering substantial benefits across diverse device capabilities.

Fengyu Gao, Ruida Zhou, Tianhao Wang et al.

Large Language Models (LLMs) rely on the contextual information embedded in examples/demonstrations to perform in-context learning (ICL). To mitigate the risk of LLMs potentially leaking private information contained in examples in the prompt, we introduce a novel data-adaptive differentially private algorithm called AdaDPSyn to generate synthetic examples from the private dataset and then use these synthetic examples to perform ICL. The objective of AdaDPSyn is to adaptively adjust the noise level in the data synthesis mechanism according to the inherent statistical properties of the data, thereby preserving high ICL accuracy while maintaining formal differential privacy guarantees. A key innovation in AdaDPSyn is the Precision-Focused Iterative Radius Reduction technique, which dynamically refines the aggregation radius - the scope of data grouping for noise addition - based on patterns observed in data clustering, thereby minimizing the amount of additive noise. We conduct extensive experiments on standard benchmarks and compare AdaDPSyn with DP few-shot generation algorithm (Tang et al., 2023). The experiments demonstrate that AdaDPSyn not only outperforms DP few-shot generation, but also maintains high accuracy levels close to those of non-private baselines, providing an effective solution for ICL with privacy protection.

Renpu Liu, Ruida Zhou, Cong Shen et al.

An intriguing property of the Transformer is its ability to perform in-context learning (ICL), where the Transformer can solve different inference tasks without parameter updating based on the contextual information provided by the corresponding input-output demonstration pairs. It has been theoretically proved that ICL is enabled by the capability of Transformers to perform gradient-descent algorithms (Von Oswald et al., 2023a; Bai et al., 2024). This work takes a step further and shows that Transformers can perform learning-to-optimize (L2O) algorithms. Specifically, for the ICL sparse recovery (formulated as LASSO) tasks, we show that a K-layer Transformer can perform an L2O algorithm with a provable convergence rate linear in K. This provides a new perspective explaining the superior ICL capability of Transformers, even with only a few layers, which cannot be achieved by the standard gradient-descent algorithms. Moreover, unlike the conventional L2O algorithms that require the measurement matrix involved in training to match that in testing, the trained Transformer is able to solve sparse recovery problems generated with different measurement matrices. Besides, Transformers as an L2O algorithm can leverage structural information embedded in the training tasks to accelerate its convergence during ICL, and generalize across different lengths of demonstration pairs, where conventional L2O algorithms typically struggle or fail. Such theoretical findings are supported by our experimental results.

Zihao Peng, Xijun Wang, Shengbo Chen et al.

Diffusion models are powerful generative models that can produce highly realistic samples for various tasks. Typically, these models are constructed using centralized, independently and identically distributed (IID) training data. However, in practical scenarios, data is often distributed across multiple clients and frequently manifests non-IID characteristics. Federated Learning (FL) can leverage this distributed data to train diffusion models, but the performance of existing FL methods is unsatisfactory in non-IID scenarios. To address this, we propose FedDDPM-Federated Learning with Denoising Diffusion Probabilistic Models, which leverages the data generative capability of diffusion models to facilitate model training. In particular, the server uses well-trained local diffusion models uploaded by each client before FL training to generate auxiliary data that can approximately represent the global data distribution. Following each round of model aggregation, the server further optimizes the global model using the auxiliary dataset to alleviate the impact of heterogeneous data on model performance. We provide a rigorous convergence analysis of FedDDPM and propose an enhanced algorithm, FedDDPM+, to reduce training overheads. FedDDPM+ detects instances of slow model learning and performs a one-shot correction using the auxiliary dataset. Experimental results validate that our proposed algorithms outperform the state-of-the-art FL algorithms on the MNIST, CIFAR10 and CIFAR100 datasets.

Kun Yang, Jing Yang, Cong Shen

In this paper, we address a crucial but often overlooked issue in applying reinforcement learning (RL) to radio resource management (RRM) in wireless communications: the mismatch between the discounted reward RL formulation and the undiscounted goal of wireless network optimization. To the best of our knowledge, we are the first to systematically investigate this discrepancy, starting with a discussion of the problem formulation followed by simulations that quantify the extent of the gap. To bridge this gap, we introduce the use of average reward RL, a method that aligns more closely with the long-term objectives of RRM. We propose a new method called the Average Reward Off policy Soft Actor Critic (ARO SAC) is an adaptation of the well known Soft Actor Critic algorithm in the average reward framework. This new method achieves significant performance improvement our simulation results demonstrate a 15% gain in the system performance over the traditional discounted reward RL approach, underscoring the potential of average reward RL in enhancing the efficiency and effectiveness of wireless network optimization.

Li Fan, Jing Yang, Cong Shen

Pre-trained Transformers, through in-context learning (ICL), have demonstrated exceptional capabilities to adapt to new tasks using example prompts \textit{without model update}. Transformer-based wireless receivers, where prompts consist of the pilot data in the form of transmitted and received signal pairs, have shown high estimation accuracy when pilot data are abundant. However, pilot information is often costly and limited in practice. In this work, we propose the \underline{DE}cision \underline{F}eedback \underline{IN}-Cont\underline{E}xt \underline{D}etection (DEFINED) solution as a new wireless receiver design, which bypasses channel estimation and directly performs symbol detection using the (sometimes extremely) limited pilot data. The key innovation in DEFINED is the proposed decision feedback mechanism in ICL, where we sequentially incorporate the detected symbols into the prompts to improve the detections for subsequent symbols. Extensive experiments across a broad range of wireless communication settings demonstrate that DEFINED achieves significant performance improvements, in some cases only needing a single pilot pair.

Yujia Mu, Xizixiang Wei, Cong Shen

Wireless federated learning (FL) relies on efficient uplink communications to aggregate model updates across distributed edge devices. Over-the-air computation (a.k.a. AirComp) has emerged as a promising approach for addressing the scalability challenge of FL over wireless links with limited communication resources. Unlike conventional methods, AirComp allows multiple edge devices to transmit uplink signals simultaneously, enabling the parameter server to directly decode the average global model. However, existing AirComp solutions are intrinsically analog, while modern wireless systems predominantly adopt digital modulations. Consequently, careful constellation designs are necessary to accurately decode the sum model updates without ambiguity. In this paper, we propose an end-to-end communication system supporting AirComp with digital modulation, aiming to overcome the challenges associated with accurate decoding of the sum signal with constellation designs. We leverage autoencoder network structures and explore the joint optimization of transmitter and receiver components. Our approach fills an important gap in the context of accurately decoding the sum signal in digital modulation-based AirComp, which can advance the deployment of FL in contemporary wireless systems.

Renpu Liu, Peng Wang, Donghao Li et al.

Reinforcement Learning from Human Feedback (RLHF) has emerged as a pivotal technique for aligning artificial intelligence systems with human values, achieving remarkable success in fine-tuning large language models. However, existing RLHF frameworks often assume that human preferences are relatively homogeneous and can be captured by a single, unified reward model. This assumption overlooks the inherent diversity and heterogeneity across individuals, limiting the adaptability of RLHF to personalized scenarios and risking misalignments that can diminish user satisfaction and trust in AI systems. In this paper, we address these challenges by introducing Low-Rank Adaptation (LoRA) into the personalized RLHF framework. We apply LoRA in the the aggregated parameter space of all personalized reward functions, thereby enabling efficient learning of personalized reward models from potentially limited local datasets. Our approach exploits potential shared structures among the local ground-truth reward models while allowing for individual adaptation, without relying on restrictive assumptions about shared representations as in prior works. We further establish sample complexity guarantees for our method. Theoretical analysis demonstrates the effectiveness of the proposed approach in capturing both shared and individual-specific structures within heterogeneous human preferences, addressing the dual challenge of personalization requirements and practical data constraints. Experimental results on real-world datasets corroborate the efficiency of our algorithm in the personalized RLHF setting.

Zihan Chen, Song Wang, Xingbo Fu et al.

The capability of in-context learning (ICL) enables large language models (LLMs) to perform novel tasks without parameter updates by conditioning on a few input-output examples. However, collecting high-quality examples for new or challenging tasks can be costly and labor-intensive. In this work, we propose a cost-efficient two-stage pipeline that reduces reliance on LLMs for data labeling. Our approach first leverages readily available cross-task examples to prompt an LLM and pseudo-label a small set of target task instances. We then introduce a graph-based label propagation method that spreads label information to the remaining target examples without additional LLM queries. The resulting fully pseudo-labeled dataset is used to construct in-task demonstrations for ICL. This pipeline combines the flexibility of cross-task supervision with the scalability of LLM-free propagation. Experiments across five tasks demonstrate that our method achieves strong performance while lowering labeling costs.

Yujia Mu, Cong Shen

Federated learning (FL) is one of the popular distributed machine learning (ML) solutions but incurs significant communication and computation costs at edge devices. Federated split learning (FSL) can train sub-models in parallel and reduce the computational burden of edge devices by splitting the model architecture. However, it still requires a high communication overhead due to transmitting the smashed data and gradients between clients and the server in every global round. Furthermore, the server must maintain separate partial models for every client, leading to a significant storage requirement. To address these challenges, this paper proposes a novel communication and storage efficient federated split learning method, termed CSE-FSL, which utilizes an auxiliary network to locally update the weights of the clients while keeping a single model at the server, hence avoiding frequent transmissions of gradients from the server and greatly reducing the storage requirement of the server. Additionally, a new model update method of transmitting the smashed data in selected epochs can reduce the amount of smashed data sent from the clients. We provide a theoretical analysis of CSE-FSL, rigorously guaranteeing its convergence under non-convex loss functions. The extensive experimental results further indicate that CSE-FSL achieves a significant communication reduction over existing FSL solutions using real-world FL tasks.

Xingbo Fu, Zehong Wang, Zihan Chen et al.

Graph learning models have demonstrated great prowess in learning expressive representations from large-scale graph data in a wide variety of real-world scenarios. As a prevalent strategy for training powerful graph learning models, the "pre-training, adaptation" scheme first pre-trains graph learning models on unlabeled graph data in a self-supervised manner and then adapts them to specific downstream tasks. During the adaptation phase, graph prompting emerges as a promising approach that learns trainable prompts while keeping the pre-trained graph learning models unchanged. In this paper, we present a systematic review of recent advancements in graph prompting. First, we introduce representative graph pre-training methods that serve as the foundation step of graph prompting. Next, we review mainstream techniques in graph prompting and elaborate on how they design learnable prompts for graph prompting. Furthermore, we summarize the real-world applications of graph prompting from different domains. Finally, we discuss several open challenges in existing studies with promising future directions in this field.

Wei Shen, Zhang Yaxiang, Minhui Huang et al.

With increasing size of large language models (LLMs), full-parameter fine-tuning imposes substantial memory demands. To alleviate this, we propose a novel memory-efficient training paradigm called Momentum Low-rank compression (MLorc). The key idea of MLorc is to compress and reconstruct the momentum of matrix parameters during training to reduce memory consumption. Compared to LoRA, MLorc avoids enforcing a fixed-rank constraint on weight update matrices and thus enables full-parameter learning. Compared to GaLore, MLorc directly compress the momentum rather than gradients, thereby better preserving the training dynamics of full-parameter fine-tuning. We provide a theoretical guarantee for its convergence under mild assumptions. Empirically, MLorc consistently outperforms other memory-efficient training methods, matches or even exceeds the performance of full fine-tuning at small ranks (e.g., $r=4$), and generalizes well across different optimizers -- all while not compromising time or memory efficiency.

Guancen Lin, Cong Shen, Aijing Lin

Time series analysis has gained significant attention due to its critical applications in diverse fields such as healthcare, finance, and sensor networks. The complexity and non-stationarity of time series make it challenging to capture the interaction patterns across different timestamps. Current approaches struggle to model higher-order interactions within time series, and focus on learning temporal or spatial dependencies separately, which limits performance in downstream tasks. To address these gaps, we propose Higher-order Cross-structural Embedding Model for Time Series (High-TS), a novel framework that jointly models both temporal and spatial perspectives by combining multiscale Transformer with Topological Deep Learning (TDL). Meanwhile, High-TS utilizes contrastive learning to integrate these two structures for generating robust and discriminative representations. Extensive experiments show that High-TS outperforms state-of-the-art methods in various time series tasks and demonstrate the importance of higher-order cross-structural information in improving model performance.

Di Wu, Chengshuai Shi, Jing Yang et al.

Reinforcement Learning from Human Feedback (RLHF) has emerged as a key technique for post-training large language models. Despite its empirical success, the theoretical understanding of RLHF is still limited, as learning the KL-regularized target with only preference feedback poses additional challenges compared with canonical RL. Existing works mostly study the reward-based Bradley-Terry (BT) preference model, and extend classical designs utilizing optimism or pessimism. This work, instead, considers the general preference model (whose practical relevance has been observed recently) and obtains performance guarantees with major, order-wise improvements over existing ones. Surprisingly, these results are derived from algorithms that directly use the empirical estimates (i.e., greedy sampling), as opposed to constructing optimistic or pessimistic estimates in previous works. This insight has a deep root in the unique structural property of the optimal policy class under the KL-regularized target, and we further specialize it to the BT model, highlighting the surprising sufficiency of greedy sampling in RLHF.

Li Fan, Peng Wang, Jing Yang et al.

Transformers have shown potential in solving wireless communication problems, particularly via in-context learning (ICL), where models adapt to new tasks through prompts without requiring model updates. However, prior ICL-based Transformer models rely on deep architectures with many layers to achieve satisfactory performance, resulting in substantial storage and computational costs. In this work, we propose CHain Of thOught Symbol dEtection (CHOOSE), a CoT-enhanced shallow Transformer framework for wireless symbol detection. By introducing autoregressive latent reasoning steps within the hidden space, CHOOSE significantly improves the reasoning capacity of shallow models (1-2 layers) without increasing model depth. This design enables lightweight Transformers to achieve detection performance comparable to much deeper models, making them well-suited for deployment on resource-constrained mobile devices. Experimental results demonstrate that our approach outperforms conventional shallow Transformers and achieves performance comparable to that of deep Transformers, while maintaining storage and computational efficiency. This represents a promising direction for implementing Transformer-based algorithms in wireless receivers with limited computational resources.

Wei Shen, Ruichuan Huang, Minhui Huang et al.

The majority of parameters in neural networks are naturally represented as matrices. However, most commonly used optimizers treat these matrix parameters as flattened vectors during optimization, potentially overlooking their inherent structural properties. Recently, an optimizer called Muon has been proposed, specifically designed to optimize matrix-structured parameters. Extensive empirical evidence shows that Muon can significantly outperform traditional optimizers when training neural networks. Nonetheless, the theoretical understanding of Muon's convergence behavior and the reasons behind its superior performance remain limited. In this work, we present a comprehensive convergence rate analysis of Muon and its comparison with Gradient Descent (GD). We further characterize the conditions under which Muon can outperform GD. Our theoretical results reveal that Muon can benefit from the low-rank and approximate blockwise diagonal structure of Hessian matrices -- phenomena widely observed in practical neural network training. Our experimental results support and corroborate the theoretical findings.