Can Chen

Can Chen, Jingbo Zhou, Fan Wang et al.

Protein representation learning methods have shown great potential to yield useful representation for many downstream tasks, especially on protein classification. Moreover, a few recent studies have shown great promise in addressing insufficient labels of proteins with self-supervised learning methods. However, existing protein language models are usually pretrained on protein sequences without considering the important protein structural information. To this end, we propose a novel structure-aware protein self-supervised learning method to effectively capture structural information of proteins. In particular, a well-designed graph neural network (GNN) model is pretrained to preserve the protein structural information with self-supervised tasks from a pairwise residue distance perspective and a dihedral angle perspective, respectively. Furthermore, we propose to leverage the available protein language model pretrained on protein sequences to enhance the self-supervised learning. Specifically, we identify the relation between the sequential information in the protein language model and the structural information in the specially designed GNN model via a novel pseudo bi-level optimization scheme. Experiments on several supervised downstream tasks verify the effectiveness of our proposed method.The code of the proposed method is available in \url{https://github.com/GGchen1997/STEPS_Bioinformatics}.

Can Chen, Xi Chen, Chen Ma et al.

Bi-level optimization, especially the gradient-based category, has been widely used in the deep learning community including hyperparameter optimization and meta-knowledge extraction. Bi-level optimization embeds one problem within another and the gradient-based category solves the outer-level task by computing the hypergradient, which is much more efficient than classical methods such as the evolutionary algorithm. In this survey, we first give a formal definition of the gradient-based bi-level optimization. Next, we delineate criteria to determine if a research problem is apt for bi-level optimization and provide a practical guide on structuring such problems into a bi-level optimization framework, a feature particularly beneficial for those new to this domain. More specifically, there are two formulations: the single-task formulation to optimize hyperparameters such as regularization parameters and the distilled data, and the multi-task formulation to extract meta-knowledge such as the model initialization. With a bi-level formulation, we then discuss four bi-level optimization solvers to update the outer variable including explicit gradient update, proxy update, implicit function update, and closed-form update. Finally, we wrap up the survey by highlighting two prospective future directions: (1) Effective Data Optimization for Science examined through the lens of task formulation. (2) Accurate Explicit Proxy Update analyzed from an optimization standpoint.

Ren Wang, Yuxuan Li, Can Chen et al.

Adversarial robustness is a critical measure of a neural network's ability to withstand adversarial attacks at inference time. While robust training techniques have improved defenses against individual $\ell_p$-norm attacks (e.g., $\ell_2$ or $\ell_\infty$), models remain vulnerable to diversified $\ell_p$ perturbations. To address this challenge, we propose a novel Robust Mode Connectivity (RMC)-oriented adversarial defense framework comprising two population-based learning phases. In Phase I, RMC searches the parameter space between two pre-trained models to construct a continuous path containing models with high robustness against multiple $\ell_p$ attacks. To improve efficiency, we introduce a Self-Robust Mode Connectivity (SRMC) module that accelerates endpoint generation in RMC. Building on RMC, Phase II presents RMC-based optimization, where RMC modules are composed to further enhance diversified robustness. To increase Phase II efficiency, we propose Efficient Robust Mode Connectivity (ERMC), which leverages $\ell_1$- and $\ell_\infty$-adversarially trained models to achieve robustness across a broad range of $p$-norms. An ensemble strategy is employed to further boost ERMC's performance. Extensive experiments across diverse datasets and architectures demonstrate that our methods significantly improve robustness against $\ell_\infty$, $\ell_2$, $\ell_1$, and hybrid attacks. Code is available at https://github.com/wangren09/MCGR.

Can Chen, Scott T. Weiss, Yang-Yu Liu

Feature selection is a powerful dimension reduction technique which selects a subset of relevant features for model construction. Numerous feature selection methods have been proposed, but most of them fail under the high-dimensional and low-sample size (HDLSS) setting due to the challenge of overfitting. In this paper, we present a deep learning-based method - GRAph Convolutional nEtwork feature Selector (GRACES) - to select important features for HDLSS data. We demonstrate empirical evidence that GRACES outperforms other feature selection methods on both synthetic and real-world datasets.

Shangyu Wu, Ying Xiong, Yufei Cui et al.

Large language models (LLMs) have demonstrated great success in various fields, benefiting from their huge amount of parameters that store knowledge. However, LLMs still suffer from several key issues, such as hallucination problems, knowledge update issues, and lacking domain-specific expertise. The appearance of retrieval-augmented generation (RAG), which leverages an external knowledge database to augment LLMs, makes up those drawbacks of LLMs. This paper reviews all significant techniques of RAG, especially in the retriever and the retrieval fusions. Besides, tutorial codes are provided for implementing the representative techniques in RAG. This paper further discusses the RAG update, including RAG with/without knowledge update. Then, we introduce RAG evaluation and benchmarking, as well as the application of RAG in representative NLP tasks and industrial scenarios. Finally, this paper discusses RAG's future directions and challenges for promoting this field's development.

Can Chen, Yang-Yu Liu

As a natural extension of link prediction on graphs, hyperlink prediction aims for the inference of missing hyperlinks in hypergraphs, where a hyperlink can connect more than two nodes. Hyperlink prediction has applications in a wide range of systems, from chemical reaction networks, social communication networks, to protein-protein interaction networks. In this paper, we provide a systematic and comprehensive survey on hyperlink prediction. We propose a new taxonomy to classify existing hyperlink prediction methods into four categories: similarity-based, probability-based, matrix optimization-based, and deep learning-based methods. To compare the performance of methods from different categories, we perform a benchmark study on various hypergraph applications using representative methods from each category. Notably, deep learning-based methods prevail over other methods in hyperlink prediction.

Hyo Seo Kim, Gang Luo, Can Chen et al.

Evolutionary algorithms for adversarial attacks leverage population-based search to discover perturbations without gradient information, but suffer from inefficient crossover operations that destroy adversarial properties through discrete interpolation. We introduce Mode Connectivity Evolutionary Attack (MoCo-EA), which replaces traditional crossover with a novel Bézier crossover operator that optimizes perturbations along a continuous Bézier curve between parent perturbations. Our key insight is that adversarial examples lie on connected manifolds where intermediate points maintain and often enhance attack effectiveness. We demonstrate three findings: (1) Successful adversarial perturbations exhibit mode connectivity; (2) Intermediate points along optimized paths achieve higher transferability than endpoints; (3) Bézier crossover dramatically outperforms discrete genetic operations while reducing convergence time and query requirements. By exploiting the geometric structure of adversarial space through path optimization, MoCo-EA provides an efficient and reliable method. Our work challenges the traditional view of adversarial examples as isolated points and opens new directions for both attack generation and defense research.

Zipeng Sun, Can Chen, Ye Yuan et al.

We study offline black-box optimization (BBO), aiming to discover improved designs from an offline dataset of designs and labels, a problem common in robotics, DNA, and materials science with limited labeled samples. While recent work applies autoregressive LLMs to BBO by formatting tasks as natural-language prompts, their left-to-right design generation struggles to capture the strong bidirectional dependencies inherent in design problems. To address this, we propose adapting diffusion LLMs to offline BBO to leverage their bidirectional modeling capabilities. However, a domain gap exists between the natural text pre-training of diffusion LLMs and the heterogeneous signals in BBO (prompts, designs, and labels). To bridge this gap, we construct a unified prompt-response corpus and introduce delimiter tokens to explicitly mark field boundaries for domain adaptation. We further propose a two-stage post-training framework to align the diffusion LLM generation with high-label designs. The first stage performs supervised fine-tuning on the unified dataset via masked-response prediction, and the second stage adopts reinforcement learning with rewards defined by label improvements. Our method achieves state-of-the-art results on Design-Bench small-data settings.

Xiuyuan Hu, Guoqing Liu, Can Chen et al. · tsinghua

Structure-based drug discovery, encompassing the tasks of protein-ligand docking and pocket-aware 3D drug design, represents a core challenge in drug discovery. However, no existing work can deal with both tasks to effectively leverage the duality between them, and current methods for each task are hindered by challenges in modeling 3D information and the limitations of available data. To address these issues, we propose 3DMolFormer, a unified dual-channel transformer-based framework applicable to both docking and 3D drug design tasks, which exploits their duality by utilizing docking functionalities within the drug design process. Specifically, we represent 3D pocket-ligand complexes using parallel sequences of discrete tokens and continuous numbers, and we design a corresponding dual-channel transformer model to handle this format, thereby overcoming the challenges of 3D information modeling. Additionally, we alleviate data limitations through large-scale pre-training on a mixed dataset, followed by supervised and reinforcement learning fine-tuning techniques respectively tailored for the two tasks. Experimental results demonstrate that 3DMolFormer outperforms previous approaches in both protein-ligand docking and pocket-aware 3D drug design, highlighting its promising application in structure-based drug discovery. The code is available at: https://github.com/HXYfighter/3DMolFormer .

Can Chen, Karla-Luise Herpoldt, Chenchao Zhao et al.

Antibodies are widely used as therapeutics, but their development requires costly affinity maturation, involving iterative mutations to enhance binding affinity.This paper explores a sequence-only scenario for affinity maturation, using solely antibody and antigen sequences. Recently AlphaFlow wraps AlphaFold within flow matching to generate diverse protein structures, enabling a sequence-conditioned generative model of structure. Building on this, we propose an alternating optimization framework that (1) fixes the sequence to guide structure generation toward high binding affinity using a structure-based affinity predictor, then (2) applies inverse folding to create sequence mutations, refined by a sequence-based affinity predictor for post selection. A key challenge is the lack of labeled data for training both predictors. To address this, we develop a co-teaching module that incorporates valuable information from noisy biophysical energies into predictor refinement. The sequence-based predictor selects consensus samples to teach the structure-based predictor, and vice versa. Our method, AffinityFlow, achieves state-of-the-art performance in affinity maturation experiments. We plan to open-source our code after acceptance.

Yehui Tang, Xiaosong Li, Fangcheng Liu et al.

The surgence of Mixture of Experts (MoE) in Large Language Models promises a small price of execution cost for a much larger model parameter count and learning capacity, because only a small fraction of parameters are activated for each input token. However, it is commonly observed that some experts are activated far more often than others, leading to system inefficiency when running the experts on different devices in parallel. Therefore, we introduce Mixture of Grouped Experts (MoGE), which groups the experts during selection and balances the expert workload better than MoE in nature. It constrains tokens to activate an equal number of experts within each predefined expert group. When a model execution is distributed on multiple devices, this architectural design ensures a balanced computational load across devices, significantly enhancing throughput, particularly for the inference phase. Further, we build Pangu Pro MoE on Ascend NPUs, a sparse model based on MoGE with 72 billion total parameters, 16 billion of which are activated for each token. The configuration of Pangu Pro MoE is optimized for Ascend 300I Duo and 800I A2 through extensive system simulation studies. Our experiments indicate that MoGE indeed leads to better expert load balancing and more efficient execution for both model training and inference on Ascend NPUs. The inference performance of Pangu Pro MoE achieves 1148 tokens/s per card and can be further improved to 1528 tokens/s per card by speculative acceleration, outperforming comparable 32B and 72B Dense models. Furthermore, we achieve an excellent cost-to-performance ratio for model inference on Ascend 300I Duo. Our studies show that Ascend NPUs are capable of training Pangu Pro MoE with massive parallelization to make it a leading model within the sub-100B total parameter class, outperforming prominent open-source models like GLM-Z1-32B and Qwen3-32B.

Zhaoying Wang, Yingdan Shi, Xiang Liu et al.

Drug-side effect research is vital for understanding adverse reactions arising in complex multi-drug therapies. However, the scarcity of higher-order datasets that capture the combinatorial effects of multiple drugs severely limits progress in this field. Existing resources such as TWOSIDES primarily focus on pairwise interactions. To fill this critical gap, we introduce HODDI, the first Higher-Order Drug-Drug Interaction Dataset, constructed from U.S. Food and Drug Administration (FDA) Adverse Event Reporting System (FAERS) records spanning the past decade, to advance computational pharmacovigilance. HODDI contains 109,744 records involving 2,506 unique drugs and 4,569 unique side effects, specifically curated to capture multi-drug interactions and their collective impact on adverse effects. Comprehensive statistical analyses demonstrate HODDI's extensive coverage and robust analytical metrics, making it a valuable resource for studying higher-order drug relationships. Evaluating HODDI with multiple models, we found that simple Multi-Layer Perceptron (MLP) can outperform graph models, while hypergraph models demonstrate superior performance in capturing complex multi-drug interactions, further validating HODDI's effectiveness. Our findings highlight the inherent value of higher-order information in drug-side effect prediction and position HODDI as a benchmark dataset for advancing research in pharmacovigilance, drug safety, and personalized medicine. The dataset and codes are available at https://github.com/TIML-Group/HODDI.

Yonghan Yang, Ye Yuan, Zipeng Sun et al.

Offline black-box optimization aims to discover novel designs with high property scores using only a static dataset, a task fundamentally challenged by the out-of-distribution (OOD) extrapolation problem. Existing approaches typically bifurcate into inverse methods, which struggle with the ill-posed nature of mapping scores to designs, and forward methods, which often lack the distributional expressivity to quantify uncertainty effectively. In this work, we propose SPADE (Support-Proximity Augmented Diffusion Estimation), a novel framework that reimagines forward surrogate modeling through the lens of conditional generative modeling. SPADE models the forward likelihood p(y|x) using a diffusion model, but with two critical enhancements to tailor it for optimization: (1) a Calibrated Diffusion Estimation module that enforces global consistency in statistical moments and pairwise rankings, and (2) a Support-Proximity Regularization mechanism that implicitly internalizes the data manifold constraint p(x) via kNN-based density estimation. Theoretically, we prove that our regularization is first-order equivalent to maximizing a Bayesian posterior with a valid design prior. Empirically, SPADE achieves state-of-the-art performance across Design-Bench tasks and an LLM data mixture optimization benchmark.

Hao Zhou, Chengming Hu, Ye Yuan et al.

Large language models (LLMs) have received considerable attention recently due to their outstanding comprehension and reasoning capabilities, leading to great progress in many fields. The advancement of LLM techniques also offers promising opportunities to automate many tasks in the telecommunication (telecom) field. After pre-training and fine-tuning, LLMs can perform diverse downstream tasks based on human instructions, paving the way to artificial general intelligence (AGI)-enabled 6G. Given the great potential of LLM technologies, this work aims to provide a comprehensive overview of LLM-enabled telecom networks. In particular, we first present LLM fundamentals, including model architecture, pre-training, fine-tuning, inference and utilization, model evaluation, and telecom deployment. Then, we introduce LLM-enabled key techniques and telecom applications in terms of generation, classification, optimization, and prediction problems. Specifically, the LLM-enabled generation applications include telecom domain knowledge, code, and network configuration generation. After that, the LLM-based classification applications involve network security, text, image, and traffic classification problems. Moreover, multiple LLM-enabled optimization techniques are introduced, such as automated reward function design for reinforcement learning and verbal reinforcement learning. Furthermore, for LLM-aided prediction problems, we discussed time-series prediction models and multi-modality prediction problems for telecom. Finally, we highlight the challenges and identify the future directions of LLM-enabled telecom networks.

Xin Mao, Joshua Pickard, Can Chen

Homogeneous polynomial dynamical systems (HPDSs), which can be equivalently represented by tensors, are essential for modeling higher-order networked systems, including ecological networks, chemical reactions, and multi-agent robotic systems. However, identifying such systems from data is challenging due to the rapid growth in the number of parameters with increasing system dimension and polynomial degree. In this article, we adopt compact and scalable representations of HPDSs leveraging low-rank tensor decompositions, including tensor train, hierarchical Tucker, and canonical polyadic decompositions. These representations exploit the intrinsic multilinear structure of HPDSs and substantially reduce the dimensionality of the parameter space. Rather than identifying the full dynamic tensor, we develop a data-driven framework that directly learns the underlying factor tensors or matrices in the associated decompositions from time-series data. The resulting identification problem is solved using alternating least-squares algorithms tailored to each tensor decomposition, achieving both accuracy and computational efficiency. We further analyze the robustness of the proposed framework in the presence of measurement noise and characterize data informativity. Finally, we demonstrate the effectiveness of our framework with numerical examples.

Xinyu Liu, Xu Zhang, Can Chen et al.

Understanding how backdoor data influences neural network training dynamics remains a complex and underexplored challenge. In this paper, we present a rigorous analysis of the impact of backdoor data on the learning process, with a particular focus on the distinct behaviors between the target class and other clean classes. Leveraging the Information Bottleneck (IB) principle connected with clustering of internal representation, We find that backdoor attacks create unique mutual information (MI) signatures, which evolve across training phases and differ based on the attack mechanism. Our analysis uncovers a surprising trade-off: visually conspicuous attacks like BadNets can achieve high stealthiness from an information-theoretic perspective, integrating more seamlessly into the model than many visually imperceptible attacks. Building on these insights, we propose a novel, dynamics-based stealthiness metric that quantifies an attack's integration at the model level. We validate our findings and the proposed metric across multiple datasets and diverse attack types, offering a new dimension for understanding and evaluating backdoor threats. Our code is available in: https://github.com/XinyuLiu71/Information_Bottleneck_Backdoor.git.

Can Chen, Shuhao Zheng, Xi Chen et al.

Label noise and class imbalance are two major issues coexisting in real-world datasets. To alleviate the two issues, state-of-the-art methods reweight each instance by leveraging a small amount of clean and unbiased data. Yet, these methods overlook class-level information within each instance, which can be further utilized to improve performance. To this end, in this paper, we propose Generalized Data Weighting (GDW) to simultaneously mitigate label noise and class imbalance by manipulating gradients at the class level. To be specific, GDW unrolls the loss gradient to class-level gradients by the chain rule and reweights the flow of each gradient separately. In this way, GDW achieves remarkable performance improvement on both issues. Aside from the performance gain, GDW efficiently obtains class-level weights without introducing any extra computational cost compared with instance weighting methods. Specifically, GDW performs a gradient descent step on class-level weights, which only relies on intermediate gradients. Extensive experiments in various settings verify the effectiveness of GDW. For example, GDW outperforms state-of-the-art methods by $2.56\%$ under the $60\%$ uniform noise setting in CIFAR10. Our code is available at https://github.com/GGchen1997/GDW-NIPS2021.

Anqi Dong, Anzhi Sheng, Xin Mao et al.

Random walks are fundamental tools for analyzing complex networked systems, including social networks, biological systems, and communication infrastructures. While classical random walks focus on pairwise interactions, many real-world systems exhibit higher-order interactions naturally modeled by hypergraphs. Existing random walk models on hypergraphs often focus on undirected structures or do not incorporate entropy-based inference, limiting their ability to capture directional flows, uncertainty, or information diffusion in complex systems. In this article, we develop a maximum-entropy random walk framework on directed hypergraphs with two interaction mechanisms: broadcasting where a pivot node activates multiple receiver nodes and merging where multiple pivot nodes jointly influence a receiver node. We infer a transition kernel via a Kullback--Leibler divergence projection onto constraints enforcing stochasticity and stationarity. The resulting optimality conditions yield a multiplicative scaling form, implemented using Sinkhorn--Schrödinger-type iterations with tensor contractions. We further analyze ergodicity, including projected linear kernels for broadcasting and tensor spectral criteria for polynomial dynamics in merging. The effectiveness of our framework is demonstrated with both synthetic and real-world examples.

Joshua Pickard, Xin Mao, Can Chen

Controlling real-world networked systems, including ecological, biomedical, and engineered networks that exhibit higher-order interactions, remains challenging due to inherent nonlinearities and large system scales. Despite extensive studies on graph controllability, the controllability properties of hypergraphs remain largely underdeveloped. Existing results focus primarily on exact controllability, which is often impractical for large-scale hypergraphs. In this article, we develop a structural controllability framework for hypergraphs by modeling hypergraph dynamics as polynomial dynamical systems. In particular, we extend classical notions of accessibility and dilation from linear graph-based systems to polynomial hypergraph dynamics and establish a hypergraph-based criterion under which the topology guarantees satisfaction of classical Lie-algebraic and Kalman-type rank conditions for almost all parameter choices. We further derive a topology-based lower bound on the minimum number of driver nodes required for structural controllability and leverage this bound to design a scalable driver node selection algorithm combining dilation-aware initialization via maximum matching with greedy accessibility expansion. We demonstrate the effectiveness and scalability of the proposed framework through numerical experiments on hypergraphs with tens to thousands of nodes and higher-order interactions.

Zhexi Gu, Jiaxin Ying, Xu-Wen Wang et al.

Accurate prediction of physician referral links is essential for optimizing care coordination and reducing fragmentation in healthcare delivery. However, existing computational methods, ranging from triadic closure heuristics to graph neural networks, fail to capture the intrinsic properties of physician referral networks, including sparsity, disassortative degree mixing, and hub-dominated topology. Here, we propose H3, a healthcare three-hop index that addresses these limitations by modeling indirect referral pathways through intermediate physicians, with degree-based normalization and a redundancy penalty to mitigate hub-mediated noise. Using Medicare Physician Shared Patient Patterns data, we evaluate H3 under two complementary prediction regimes: within-period prediction, which assesses recovery of contemporaneous referral links under sparse conditions, and cross-period prediction, which tests robustness to temporal shift as referral windows expand. Across both regimes, H3 consistently outperforms classical heuristics and deep learning-based baselines. Unlike black-box neural network approaches, H3 produces fully decomposable predictions traceable to specific intermediary physicians, offering a transparent and deployable solution for referral network completion.

Dev Mistry, Feng Qiu, Bo Chen et al.

Reliable photovoltaic defect identification is essential for maintaining energy yield, ensuring warranty compliance, and enabling scalable inspection of rapidly expanding solar fleets. Although recent advances in computer vision have improved automated defect detection, most existing systems operate as opaque classifiers that provide limited diagnostic insight for high-stakes energy infrastructure. Here we introduce REVL-PV, a vision-language framework that embeds domain-specific diagnostic reasoning into multimodal learning across electroluminescence, thermal, and visible-light imagery. By requiring the model to link visual evidence to plausible defect mechanisms before classification, the framework produces structured diagnostic reports aligned with professional photovoltaic inspection practice. Evaluated on 1,927 real-world modules spanning eight defect categories, REVL-PV achieves 93\% classification accuracy while producing interpretable diagnostic rationales and maintaining strong robustness under realistic image corruptions. A blind concordance study with a certified solar inspection expert shows strong semantic alignment between model explanations and expert assessments across defect identification, root-cause attribution, and visual descriptions. These results demonstrate that reasoning-aware multimodal learning establishes a general paradigm for trustworthy AI-assisted inspection of photovoltaic energy infrastructure.

Yichun Yin, Wenyong Huang, Kaikai Song et al.

We present Pangu Ultra, a Large Language Model (LLM) with 135 billion parameters and dense Transformer modules trained on Ascend Neural Processing Units (NPUs). Although the field of LLM has been witnessing unprecedented advances in pushing the scale and capability of LLM in recent years, training such a large-scale model still involves significant optimization and system challenges. To stabilize the training process, we propose depth-scaled sandwich normalization, which effectively eliminates loss spikes during the training process of deep models. We pre-train our model on 13.2 trillion diverse and high-quality tokens and further enhance its reasoning capabilities during post-training. To perform such large-scale training efficiently, we utilize 8,192 Ascend NPUs with a series of system optimizations. Evaluations on multiple diverse benchmarks indicate that Pangu Ultra significantly advances the state-of-the-art capabilities of dense LLMs such as Llama 405B and Mistral Large 2, and even achieves competitive results with DeepSeek-R1, whose sparse model structure contains much more parameters. Our exploration demonstrates that Ascend NPUs are capable of efficiently and effectively training dense models with more than 100 billion parameters. Our model and system will be available for our commercial customers.

Ye Yuan, Can Chen, Christopher Pal et al.

In offline multi-objective optimization (MOO), we leverage an offline dataset of designs and their associated labels to simultaneously minimize multiple objectives. This setting more closely mirrors complex real-world problems compared to single-objective optimization. Recent works mainly employ evolutionary algorithms and Bayesian optimization, with limited attention given to the generative modeling capabilities inherent in such data. In this study, we explore generative modeling in offline MOO through flow matching, noted for its effectiveness and efficiency. We introduce ParetoFlow, specifically designed to guide flow sampling to approximate the Pareto front. Traditional predictor (classifier) guidance is inadequate for this purpose because it models only a single objective. In response, we propose a multi-objective predictor guidance module that assigns each sample a weight vector, representing a weighted distribution across multiple objective predictions. A local filtering scheme is introduced to address non-convex Pareto fronts. These weights uniformly cover the entire objective space, effectively directing sample generation towards the Pareto front. Since distributions with similar weights tend to generate similar samples, we introduce a neighboring evolution module to foster knowledge sharing among neighboring distributions. This module generates offspring from these distributions, and selects the most promising one for the next iteration. Our method achieves state-of-the-art performance across various tasks.

Minsu Kim, Jiayao Gu, Ye Yuan et al.

Offline optimization is a fundamental challenge in science and engineering, where the goal is to optimize black-box functions using only offline datasets. This setting is particularly relevant when querying the objective function is prohibitively expensive or infeasible, with applications spanning protein engineering, material discovery, neural architecture search, and beyond. The main difficulty lies in accurately estimating the objective landscape beyond the available data, where extrapolations are fraught with significant epistemic uncertainty. This uncertainty can lead to objective hacking(reward hacking), exploiting model inaccuracies in unseen regions, or other spurious optimizations that yield misleadingly high performance estimates outside the training distribution. Recent advances in model-based optimization(MBO) have harnessed the generalization capabilities of deep neural networks to develop offline-specific surrogate and generative models. Trained with carefully designed strategies, these models are more robust against out-of-distribution issues, facilitating the discovery of improved designs. Despite its growing impact in accelerating scientific discovery, the field lacks a comprehensive review. To bridge this gap, we present the first thorough review of offline MBO. We begin by formalizing the problem for both single-objective and multi-objective settings and by reviewing recent benchmarks and evaluation metrics. We then categorize existing approaches into two key areas: surrogate modeling, which emphasizes accurate function approximation in out-of-distribution regions, and generative modeling, which explores high-dimensional design spaces to identify high-performing designs. Finally, we examine the key challenges and propose promising directions for advancement in this rapidly evolving field including safe control of superintelligent systems.

Ye Yuan, Youyuan Zhang, Can Chen et al.

Offline model-based optimization (MBO) aims to maximize a black-box objective function using only an offline dataset of designs and scores. These tasks span various domains, such as robotics, material design, and protein and molecular engineering. A common approach involves training a surrogate model using existing designs and their corresponding scores, and then generating new designs through gradient-based updates with respect to the surrogate model. This method suffers from the out-of-distribution issue, where the surrogate model may erroneously predict high scores for unseen designs. To address this challenge, we introduce a novel method, Design Editing for Offline Model-based Optimization (DEMO), which leverages a diffusion prior to calibrate overly optimized designs. DEMO first generates pseudo design candidates by performing gradient ascent with respect to a surrogate model. While these pseudo design candidates contain information beyond the offline dataset, they might be invalid or have erroneously high predicted scores. Therefore, to address this challenge while utilizing the information provided by pseudo design candidates, we propose an editing process to refine these pseudo design candidates. We introduce noise to the pseudo design candidates and subsequently denoise them with a diffusion prior trained on the offline dataset, ensuring they align with the distribution of valid designs. Empirical evaluations on seven offline MBO tasks show that, with properly tuned hyperparameters, DEMOs score is competitive with the best previously reported scores in the literature.

Yehui Tang, Yichun Yin, Yaoyuan Wang et al.

Sparse large language models (LLMs) with Mixture of Experts (MoE) and close to a trillion parameters are dominating the realm of most capable language models. However, the massive model scale poses significant challenges for the underlying software and hardware systems. In this paper, we aim to uncover a recipe to harness such scale on Ascend NPUs. The key goals are better usage of the computing resources under the dynamic sparse model structures and materializing the expected performance gain on the actual hardware. To select model configurations suitable for Ascend NPUs without repeatedly running the expensive experiments, we leverage simulation to compare the trade-off of various model hyperparameters. This study led to Pangu Ultra MoE, a sparse LLM with 718 billion parameters, and we conducted experiments on the model to verify the simulation results. On the system side, we dig into Expert Parallelism to optimize the communication between NPU devices to reduce the synchronization overhead. We also optimize the memory efficiency within the devices to further reduce the parameter and activation management overhead. In the end, we achieve an MFU of 30.0% when training Pangu Ultra MoE, with performance comparable to that of DeepSeek R1, on 6K Ascend NPUs, and demonstrate that the Ascend system is capable of harnessing all the training stages of the state-of-the-art language models. Extensive experiments indicate that our recipe can lead to efficient training of large-scale sparse language models with MoE. We also study the behaviors of such models for future reference.

Can Chen, Gabriel Oliveira, Hossein Sharifi Noghabi et al.

Time series~(TS) modeling is essential in dynamic systems like weather prediction and anomaly detection. Recent studies utilize Large Language Models (LLMs) for TS modeling, leveraging their powerful pattern recognition capabilities. These methods primarily position LLMs as the predictive backbone, often omitting the mathematical modeling within traditional TS models, such as periodicity. However, disregarding the potential of LLMs also overlooks their pattern recognition capabilities. To address this gap, we introduce \textit{LLM-TS Integrator}, a novel framework that effectively integrates the capabilities of LLMs into traditional TS modeling. Central to this integration is our \textit{mutual information} module. The core of this \textit{mutual information} module is a traditional TS model enhanced with LLM-derived insights for improved predictive abilities. This enhancement is achieved by maximizing the mutual information between traditional model's TS representations and LLM's textual representation counterparts, bridging the two modalities. Moreover, we recognize that samples vary in importance for two losses: traditional prediction and mutual information maximization. To address this variability, we introduce the \textit{sample reweighting} module to improve information utilization. This module assigns dual weights to each sample: one for prediction loss and another for mutual information loss, dynamically optimizing these weights via bi-level optimization. Our method achieves state-of-the-art or comparable performance across five mainstream TS tasks, including short-term and long-term forecasting, imputation, classification, and anomaly detection.

Xiuyuan Hu, Guoqing Liu, Can Chen et al. · tsinghua

Structure-based drug design (SBDD) is a critical task in drug discovery, requiring the generation of molecular information across two distinct modalities: discrete molecular graphs and continuous 3D coordinates. However, existing SBDD methods often overlook two key challenges: (1) the multi-modal nature of this task and (2) the causal relationship between these modalities, limiting their plausibility and performance. To address both challenges, we propose TransDiffSBDD, an integrated framework combining autoregressive transformers and diffusion models for SBDD. Specifically, the autoregressive transformer models discrete molecular information, while the diffusion model samples continuous distributions, effectively resolving the first challenge. To address the second challenge, we design a hybrid-modal sequence for protein-ligand complexes that explicitly respects the causality between modalities. Experiments on the CrossDocked2020 benchmark demonstrate that TransDiffSBDD outperforms existing baselines.

Can Chen, Jun-Kun Wang

The technique of ``testing by betting" frames nonparametric sequential hypothesis testing as a multiple-round game, where a player bets on future observations that arrive in a streaming fashion, accumulates wealth that quantifies evidence against the null hypothesis, and rejects the null once the wealth exceeds a specified threshold while controlling the false positive error. Designing an online learning algorithm that achieves a small regret in the game can help rapidly accumulate the bettor's wealth, which in turn can shorten the time to reject the null hypothesis under the alternative $H_1$. However, many of the existing works employ the Online Newton Step (ONS) to update within a halved decision space to avoid a gradient explosion issue, which is potentially conservative for rapid wealth accumulation. In this paper, we introduce a novel strategy utilizing interior-point methods in optimization that allows updates across the entire interior of the decision space without the risk of gradient explosion. Our approach not only maintains strong statistical guarantees but also facilitates faster null hypothesis rejection, while being as computationally lightweight as ONS thanks to its closed-form updates.

Yixuan Li, Can Chen, Jiajun Li et al.

By exploiting the correlation between the structure and the solution of Mixed-Integer Linear Programming (MILP), Machine Learning (ML) has become a promising method for solving large-scale MILP problems. Existing ML-based MILP solvers mainly focus on end-to-end solution learning, which suffers from the scalability issue due to the high dimensionality of the solution space. Instead of directly learning the optimal solution, this paper aims to learn a reduced and equivalent model of the original MILP as an intermediate step. The reduced model often corresponds to interpretable operations and is much simpler, enabling us to solve large-scale MILP problems much faster than existing commercial solvers. However, current approaches rely only on the optimal reduced model, overlooking the significant preference information of all reduced models. To address this issue, this paper proposes a preference-based model reduction learning method, which considers the relative performance (i.e., objective cost and constraint feasibility) of all reduced models on each MILP instance as preferences. We also introduce an attention mechanism to capture and represent preference information, which helps improve the performance of model reduction learning tasks. Moreover, we propose a SetCover based pruning method to control the number of reduced models (i.e., labels), thereby simplifying the learning process. Evaluation on real-world MILP problems shows that 1) compared to the state-of-the-art model reduction ML methods, our method obtains nearly 20% improvement on solution accuracy, and 2) compared to the commercial solver Gurobi, two to four orders of magnitude speedups are achieved.

Jiajie Song, Ningfang Song, Xiong Pan et al.

With the rapid development of urban underground rail vehicles,subway positioning, which plays a fundamental role in the traffic navigation and collision avoidance systems, has become a research hot-spot these years. Most current subway positioning methods rely on localization beacons densely pre-installed alongside the railway tracks, requiring massive costs for infrastructure and maintenance, while commonly lacking flexibility and anti-interference ability. In this paper, we propose a low-cost and real-time visual-assisted self-localization framework to address the robust and convenient positioning problem for subways. Firstly, we perform aerial view rail sleeper detection based on the fast and efficient YOLOv8n network. The detection results are then used to achieve real-time correction of mileage values combined with geometric positioning information, obtaining precise subway locations. Front camera Videos for subway driving scenes along a 6.9 km route are collected and annotated from the simulator for validation of the proposed method. Experimental results show that our aerial view sleeper detection algorithm can efficiently detect sleeper positions with F1-score of 0.929 at 1111 fps, and that the proposed positioning framework achieves a mean percentage error of 0.1\%, demonstrating its continuous and high-precision self-localization capability.

Ziqin He, Mengqi Hu, Yifei Lou et al.

Dynamic mode decomposition (DMD) has become a powerful data-driven method for analyzing the spatiotemporal dynamics of complex, high-dimensional systems. However, conventional DMD methods are limited to matrix-based formulations, which might be inefficient or inadequate for modeling inherently multidimensional data including images, videos, and higher-order networks. In this letter, we propose tensor dynamic mode decomposition (TDMD), a novel extension of DMD to third-order tensors based on the recently developed T-product framework. By incorporating tensor factorization techniques, TDMD achieves more efficient computation and better preservation of spatial and temporal structures in multiway data for tasks such as state reconstruction and dynamic component separation, compared to standard DMD with data flattening. We demonstrate the effectiveness of TDMD on both synthetic and real-world datasets.

Ding Zhang, Jane Downer, Can Chen et al.

Graph classification is essential for understanding complex biological systems, where molecular structures and interactions are naturally represented as graphs. Traditional graph neural networks (GNNs) perform well on static tasks but struggle in dynamic settings due to catastrophic forgetting. We present Perturbed and Sparsified Continual Graph Learning (PSCGL), a robust and efficient continual graph learning framework for graph data classification, specifically targeting biological datasets. We introduce a perturbed sampling strategy to identify critical data points that contribute to model learning and a motif-based graph sparsification technique to reduce storage needs while maintaining performance. Additionally, our PSCGL framework inherently defends against graph backdoor attacks, which is crucial for applications in sensitive biological contexts. Extensive experiments on biological datasets demonstrate that PSCGL not only retains knowledge across tasks but also enhances the efficiency and robustness of graph classification models in biology.

Can Chen, Jun-Kun Wang

Developing algorithms to differentiate between machine-generated texts and human-written texts has garnered substantial attention in recent years. Existing methods in this direction typically concern an offline setting where a dataset containing a mix of real and machine-generated texts is given upfront, and the task is to determine whether each sample in the dataset is from a large language model (LLM) or a human. However, in many practical scenarios, sources such as news websites, social media accounts, and online forums publish content in a streaming fashion. Therefore, in this online scenario, how to quickly and accurately determine whether the source is an LLM with strong statistical guarantees is crucial for these media or platforms to function effectively and prevent the spread of misinformation and other potential misuse of LLMs. To tackle the problem of online detection, we develop an algorithm based on the techniques of sequential hypothesis testing by betting that not only builds upon and complements existing offline detection techniques but also enjoys statistical guarantees, which include a controlled false positive rate and the expected time to correctly identify a source as an LLM. Experiments were conducted to demonstrate the effectiveness of our method.

Shiyi Yang, Can Chen, Didong Li

Networks with higher-order interactions, prevalent in biological, social, and information systems, are naturally represented as hypergraphs, yet their structural complexity poses fundamental challenges for geometric characterization. While curvature-based methods offer powerful insights in graph analysis, existing extensions to hypergraphs suffer from critical trade-offs: combinatorial approaches such as Forman-Ricci curvature capture only coarse features, whereas geometric methods like Ollivier-Ricci curvature offer richer expressivity but demand costly optimal transport computations. To address these challenges, we introduce hypergraph lower Ricci curvature (HLRC), a novel curvature metric defined in closed form that achieves a principled balance between interpretability and efficiency. Evaluated across diverse synthetic and real-world hypergraph datasets, HLRC consistently reveals meaningful higher-order organization, distinguishing intra- from inter-community hyperedges, uncovering latent semantic labels, tracking temporal dynamics, and supporting robust clustering of hypergraphs based on global structure. By unifying geometric sensitivity with algorithmic simplicity, HLRC provides a versatile foundation for hypergraph analytics, with broad implications for tasks including node classification, anomaly detection, and generative modeling in complex systems.

Can Chen, David Heurtel-Depeiges, Robert M. Vernon et al.

Protein language models (pLMs) pre-trained on vast protein sequence databases excel at various downstream tasks but lack the structural knowledge essential for many biological applications. To address this, we integrate structural insights from pre-trained protein graph neural networks (pGNNs) into pLMs through a latent-level contrastive learning task. This task aligns residue representations from pLMs with those from pGNNs across multiple proteins, enriching pLMs with inter-protein structural knowledge. Additionally, we incorporate a physical-level task that infuses intra-protein structural knowledge by optimizing pLMs to predict structural tokens. The proposed dual-task framework effectively incorporates both inter-protein and intra-protein structural knowledge into pLMs. Given the variability in the quality of protein structures in PDB, we further introduce a residue loss selection module, which uses a small model trained on high-quality structures to select reliable yet challenging residue losses for the pLM to learn. Applying our structure alignment method to the state-of-the-art ESM2 and AMPLIFY results in notable performance gains across a wide range of tasks, including a 12.7% increase in ESM2 contact prediction. The data, code, and resulting SaESM2 and SaAMPLIFY models will be released on Hugging Face.

Amit Surana, Can Chen, Indika Rajapakse

In this paper, we propose two novel approaches for hypergraph comparison. The first approach transforms the hypergraph into a graph representation for use of standard graph dissimilarity measures. The second approach exploits the mathematics of tensors to intrinsically capture multi-way relations. For each approach, we present measures that assess hypergraph dissimilarity at a specific scale or provide a more holistic multi-scale comparison. We test these measures on synthetic hypergraphs and apply them to biological datasets.

Can Chen, Luca Zanotti Fragonara, Antonios Tsourdos

Autonomous systems need to localize and track surrounding objects in 3D space for safe motion planning. As a result, 3D multi-object tracking (MOT) plays a vital role in autonomous navigation. Most MOT methods use a tracking-by-detection pipeline, which includes object detection and data association processing. However, many approaches detect objects in 2D RGB sequences for tracking, which is lack of reliability when localizing objects in 3D space. Furthermore, it is still challenging to learn discriminative features for temporally-consistent detection in different frames, and the affinity matrix is normally learned from independent object features without considering the feature interaction between detected objects in the different frames. To settle these problems, We firstly employ a joint feature extractor to fuse the 2D and 3D appearance features captured from both 2D RGB images and 3D point clouds respectively, and then propose a novel convolutional operation, named RelationConv, to better exploit the correlation between each pair of objects in the adjacent frames, and learn a deep affinity matrix for further data association. We finally provide extensive evaluation to reveal that our proposed model achieves state-of-the-art performance on KITTI tracking benchmark.

Can Chen, Luca Zanotti Fragonara, Antonios Tsourdos

When localizing and detecting 3D objects for autonomous driving scenes, obtaining information from multiple sensor (e.g. camera, LIDAR) typically increases the robustness of 3D detectors. However, the efficient and effective fusion of different features captured from LIDAR and camera is still challenging, especially due to the sparsity and irregularity of point cloud distributions. This notwithstanding, point clouds offer useful complementary information. In this paper, we would like to leverage the advantages of LIDAR and camera sensors by proposing a deep neural network architecture for the fusion and the efficient detection of 3D objects by identifying their corresponding 3D bounding boxes with orientation. In order to achieve this task, instead of densely combining the point-wise feature of the point cloud and the related pixel features, we propose a novel fusion algorithm by projecting a set of 3D Region of Interests (RoIs) from the point clouds to the 2D RoIs of the corresponding the images. Finally, we demonstrate that our deep fusion approach achieves state-of-the-art performance on the KITTI 3D object detection challenging benchmark.

Can Chen, Amit Surana, Anthony Bloch et al.

In this paper, we develop a notion of controllability for hypergraphs via tensor algebra and polynomial control theory. Inspired by uniform hypergraphs, we propose a new tensor-based multilinear dynamical system representation, and derive a Kalman-rank-like condition to determine the minimum number of control nodes (MCN) needed to achieve controllability of even uniform hypergraphs. We present an efficient heuristic to obtain the MCN. MCN can be used as a measure of robustness, and we show that it is related to the hypergraph degree distribution in simulated examples. Finally, we use MCN to examine robustness in real biological networks.

Can Chen, Indika Rajapakse

In this paper, we develop the notion of entropy for uniform hypergraphs via tensor theory. We employ the probability distribution of the generalized singular values, calculated from the higher-order singular value decomposition of the Laplacian tensors, to fit into the Shannon entropy formula. We show that this tensor entropy is an extension of von Neumann entropy for graphs. In addition, we establish results on the lower and upper bounds of the entropy and demonstrate that it is a measure of regularity for uniform hypergraphs in simulated and experimental data. We exploit the tensor train decomposition in computing the proposed tensor entropy efficiently. Finally, we introduce the notion of robustness for uniform hypergraphs.

Can Chen, Luca Zanotti Fragonara, Antonios Tsourdos

In order to achieve better performance for point cloud analysis, many researchers apply deeper neural networks using stacked Multi-Layer-Perceptron (MLP) convolutions over irregular point cloud. However, applying dense MLP convolutions over large amount of points (e.g. autonomous driving application) leads to inefficiency in memory and computation. To achieve high performance but less complexity, we propose a deep-wide neural network, called ShufflePointNet, to exploit fine-grained local features and reduce redundancy in parallel using group convolution and channel shuffle operation. Unlike conventional operation that directly applies MLPs on high-dimensional features of point cloud, our model goes wider by splitting features into groups in advance, and each group with certain smaller depth is only responsible for respective MLP operation, which can reduce complexity and allows to encode more useful information. Meanwhile, we connect communication between groups by shuffling groups in feature channel to capture fine-grained features. We claim that, multi-branch method for wider neural networks is also beneficial to feature extraction for point cloud. We present extensive experiments for shape classification task on ModelNet40 dataset and semantic segmentation task on large scale datasets ShapeNet part, S3DIS and KITTI. We further perform ablation study and compare our model to other state-of-the-art algorithms in terms of complexity and accuracy.

Can Chen, Luca Zanotti Fragonara, Antonios Tsourdos

The analyses relying on 3D point clouds are an utterly complex task, often involving million of points, but also requiring computationally efficient algorithms because of many real-time applications; e.g. autonomous vehicle. However, point clouds are intrinsically irregular and the points are sparsely distributed in a non-Euclidean space, which normally requires point-wise processing to achieve high performances. Although shared filter matrices and pooling layers in convolutional neural networks (CNNs) are capable of reducing the dimensionality of the problem and extracting high-level information simultaneously, grids and highly regular data format are required as input. In order to balance model performance and complexity, we introduce a novel neural network architecture exploiting local features from a manually subsampled point set. In our network, a recursive farthest point sampling method is firstly applied to efficiently cover the entire point set. Successively, we employ the k-nearest neighbours (knn) algorithm to gather local neighbourhood for each group of the subsampled points. Finally, a multiple layer perceptron (MLP) is applied on the subsampled points and edges that connect corresponding point and neighbours to extract local features. The architecture has been tested for both shape classification and segmentation using the ModelNet40 and ShapeNet part datasets, in order to show that the network achieves the best trade-off in terms of competitive performance when compared to other state-of-the-art algorithms.

Can Chen, Luca Zanotti Fragonara, Antonios Tsourdos

Exploiting fine-grained semantic features on point cloud is still challenging due to its irregular and sparse structure in a non-Euclidean space. Among existing studies, PointNet provides an efficient and promising approach to learn shape features directly on unordered 3D point cloud and has achieved competitive performance. However, local feature that is helpful towards better contextual learning is not considered. Meanwhile, attention mechanism shows efficiency in capturing node representation on graph-based data by attending over neighboring nodes. In this paper, we propose a novel neural network for point cloud, dubbed GAPNet, to learn local geometric representations by embedding graph attention mechanism within stacked Multi-Layer-Perceptron (MLP) layers. Firstly, we introduce a GAPLayer to learn attention features for each point by highlighting different attention weights on neighborhood. Secondly, in order to exploit sufficient features, a multi-head mechanism is employed to allow GAPLayer to aggregate different features from independent heads. Thirdly, we propose an attention pooling layer over neighbors to capture local signature aimed at enhancing network robustness. Finally, GAPNet applies stacked MLP layers to attention features and local signature to fully extract local geometric structures. The proposed GAPNet architecture is tested on the ModelNet40 and ShapeNet part datasets, and achieves state-of-the-art performance in both shape classification and part segmentation tasks.