Zhong Guan

Lei Zhang, Fubo Sun, Haipeng Yang et al.

Graph clustering is essential in graph analysis for revealing structural patterns and node communities. Despite recent advances in self-supervised contrastive learning that have improved clustering via structural and attribute signals, existing methods still struggle to flexibly capture high-order local structures and often overlook global semantics in complex graphs. These limitations lead to suboptimal node representations, especially in real-world graphs with fragmented structures and ambiguous cluster boundaries. To address these limitations, a contrastive graph clustering framework is proposed to jointly integrate multi-scale local structures with global semantics via attention mechanisms. At the local level, GNN-based topological signals extracted from multiple propagation depths are adaptively fused through attention-based weighting to capture multi-scale neighborhood features. At the global level, semantic prototypes derived from dynamically evolving cluster centers are adaptively aggregated through attention to guide node representations and enhance inter-cluster separability. The model is trained under a dual-view contrastive learning paradigm with a hybrid objective that combines instance-level and structure-aware losses to improve representation robustness and discrimination. Experiments on eight real-world graph datasets demonstrate that our method achieves competitive clustering performance. Code is available at https://github.com/vege12138/w2.

Yitian Zhang, Yonghong chen, Youming Chen et al.

Neural field surrogates can accelerate photonic design loops, but a surrogate that looks accurate in global field error can still mis-rank candidate devices when the final decision depends on localized output-port readouts. This risk is acute in propagation-dominated MMI splitters and couplers, where port power, splitting, phase, and coupling are determined by accumulated modal interference and output-window aggregation rather than by average field similarity alone. We study this field-to-design mismatch through a Field/Mediator/Readout view that separates dense complex-field error from propagation-profile and output-window errors before port aggregation. To align the surrogate with this chain, we propose PaNO, a propagation-aligned neural operator that keeps the full-field prediction interface while organizing latent states around local boundary structure, transverse modal content, axial propagation, and cross-mode interaction. We also evaluate PaNO-R2, an output-aware feedback variant for residual field components near the port region. On a 15-wavelength tunable $3{\times}3$ MMI benchmark with 4608 held-out fields, PaNO lowers NeurOLight's port-power error from 0.2018 to 0.0739 despite slightly higher cMAE, showing that global field accuracy alone is not sufficient for design-relevant readout fidelity. PaNO-R2 attains the best cMAE, propagation-profile error, output-profile error, and port-power error, reducing NeurOLight's port-power and output-profile errors by 72.7\% and 72.5\%.

Zhong Guan, Yongjian Guo, Haoran Sun et al.

Asynchronous reinforcement learning improves rollout throughput for large language model agents by decoupling sample generation from policy optimization, but it also introduces a critical failure mode for PPO-style off-policy correction. In heterogeneous training systems, the total importance ratio should ideally be decomposed into two semantically distinct factors: a \emph{training--inference discrepancy term} that aligns inference-side and training-side distributions at the same behavior-policy version, and a \emph{policy-staleness term} that constrains the update from the historical policy to the current policy. We show that practical asynchronous pipelines with delayed updates and partial rollouts often lose the required historical training-side logits, or old logits. This missing-old-logit problem entangles discrepancy repair with staleness correction, breaks the intended semantics of decoupled correction, and makes clipping and masking thresholds interact undesirably. To address this issue, we study both exact and approximate correction routes. We propose three exact old-logit acquisition strategies: snapshot-based version tracking, a dedicated old-logit model, and synchronization via partial rollout interruption, and compare their system trade-offs. From the perspective of approximate correction, we focus on preserving the benefits of decoupled correction through a more appropriate approximate policy when exact old logits cannot be recovered at low cost, without incurring extra system overhead. Following this analysis, we adopt a revised PPO-EWMA method, which achieves significant gains in both training speed and optimization performance. Code at https://github.com/millioniron/ROLL.

Chen Zhou, Haoran Sun, Hedan Yang et al.

Embodied intelligence is a key step towards Artificial General Intelligence (AGI), yet its development faces multiple challenges including data, frameworks, infrastructure, and evaluation systems. To address these issues, we have, for the first time in the industry, launched a cloud-based, thousand-GPU distributed training platform for embodied intelligence, built upon the widely adopted LeRobot framework, and have systematically overcome bottlenecks across the entire pipeline. At the data layer, we have restructured the data pipeline to optimize the flow of embodied training data. In terms of training, for the GR00T-N1.5 model, utilizing thousand-GPU clusters and data at the scale of hundreds of millions, the single-round training time has been reduced from 15 hours to just 22 minutes, achieving a 40-fold speedup. At the model layer, by combining variable-length FlashAttention and Data Packing, we have moved from sample redundancy to sequence integration, resulting in a 188% speed increase; π-0.5 attention optimization has accelerated training by 165%; and FP8 quantization has delivered a 140% speedup. On the infrastructure side, relying on high-performance storage, a 3.2T RDMA network, and a Ray-driven elastic AI data lake, we have achieved deep synergy among data, storage, communication, and computation. We have also built an end-to-end evaluation system, creating a closed loop from training to simulation to assessment. This framework has already been fully validated on thousand-GPU clusters, laying a crucial technical foundation for the development and application of next-generation autonomous intelligent robots, and is expected to accelerate the arrival of the era of human-machine integration.

Zhong Guan, Haoran Sun, Yongjian Guo et al.

In recent years, Vision-Language-Action (VLA) models have emerged as a crucial pathway towards general embodied intelligence, yet their training efficiency has become a key bottleneck. Although existing reinforcement learning (RL)-based training frameworks like RLinf can enhance model generalization, they still rely on synchronous execution, leading to severe resource underutilization and throughput limitations during environment interaction, policy generation (rollout), and model update phases (actor). To overcome this challenge, this paper, for the first time, proposes and implements a fully-asynchronous policy training framework encompassing the entire pipeline from environment interaction, rollout generation, to actor policy updates. Systematically drawing inspiration from asynchronous optimization ideas in large model RL, our framework designs a multi-level decoupled architecture. This includes asynchronous parallelization of environment interaction and trajectory collection, streaming execution for policy generation, and decoupled scheduling for training updates. We validated the effectiveness of our method across diverse VLA models and environments. On the LIBERO benchmark, the framework achieves throughput improvements of up to 59.25\% compared to existing synchronous strategies. When deeply optimizing separation strategies, throughput can be increased by as much as 126.67\%. We verified the effectiveness of each asynchronous component via ablation studies. Scaling law validation across 8 to 256 GPUs demonstrates our method's excellent scalability under most conditions.

Weilong Guan, Kaihan Yang, Yinsheng Chen et al.

Physics-informed neural networks (PINNs) have been widely applied in different fields due to their effectiveness in solving partial differential equations (PDEs). However, the accuracy and efficiency of PINNs need to be considerably improved for scientific and commercial use. To address this issue, we systematically propose a novel dimension-augmented physics-informed neural network (DaPINN), which simultaneously and significantly improves the accuracy and efficiency of the PINN. In the DaPINN model, we introduce inductive bias in the neural network to enhance network generalizability by adding a special regularization term to the loss function. Furthermore, we manipulate the network input dimension by inserting additional sample features and incorporating the expanded dimensionality in the loss function. Moreover, we verify the effectiveness of power series augmentation, Fourier series augmentation and replica augmentation, in both forward and backward problems. In most experiments, the error of DaPINN is 1$\sim$2 orders of magnitude lower than that of PINN. The results show that the DaPINN outperforms the original PINN in terms of both accuracy and efficiency with a reduced dependence on the number of sample points. We also discuss the complexity of the DaPINN and its compatibility with other methods.

Yucheng Guo, Yongjian Guo, Zhong Guan et al.

In video generation models, particularly world models, training large-scale video diffusion Transformers (such as DiT and MMDiT) poses significant computational challenges due to the extreme variance in sequence lengths within mixed-mode datasets. Existing bucket-based data loading strategies typically rely on "equal token length" constraints. This approach fails to account for the quadratic complexity of self-attention mechanisms, leading to severe load imbalance and underutilization of GPU resources. This paper proposes \textit{AdaptiveLoad}, an integrated optimization framework consisting of two core components: (1) A dual-constraint adaptive load balancing system, which eliminates long-sequence bottlenecks by simultaneously limiting memory consumption and computational load ($B \times S^p \le M_{\text{comp}}$); (2) A fused LayerNorm-Modulate CUDA kernel, which utilizes a D-tile coalesced reduction strategy to increase throughput and alleviate memory pressure. Experimental results on the Wan 2.1 world model demonstrate that our method reduces the computational imbalance rate from 39\% to 18.9\%, improves peak VRAM utilization efficiency by 22.7\%, and achieves an overall training throughput increase of 27.2\%.

Zhong Guan, Likang Wu, Hongke Zhao et al.

Attention mechanisms are critical to the success of large language models (LLMs), driving significant advancements in multiple fields. However, for graph-structured data, which requires emphasis on topological connections, they fall short compared to message-passing mechanisms on fixed links, such as those employed by Graph Neural Networks (GNNs). This raises a question: ``Does attention fail for graphs in natural language settings?'' Motivated by these observations, we embarked on an empirical study from the perspective of attention mechanisms to explore how LLMs process graph-structured data. The goal is to gain deeper insights into the attention behavior of LLMs over graph structures. We uncovered unique phenomena regarding how LLMs apply attention to graph-structured data and analyzed these findings to improve the modeling of such data by LLMs. The primary findings of our research are: 1) While LLMs can recognize graph data and capture text-node interactions, they struggle to model inter-node relationships within graph structures due to inherent architectural constraints. 2) The attention distribution of LLMs across graph nodes does not align with ideal structural patterns, indicating a failure to adapt to graph topology nuances. 3) Neither fully connected attention nor fixed connectivity is optimal; each has specific limitations in its application scenarios. Instead, intermediate-state attention windows improve LLM training performance and seamlessly transition to fully connected windows during inference. Source code: \href{https://github.com/millioniron/LLM_exploration}{LLM4Exploration}

Yucheng Guo, Yongjian Guo, Zhong Guan et al.

The rapid evolution of Embodied AI has enabled Vision-Language-Action (VLA) models to excel in multimodal perception and task execution. However, applying Reinforcement Learning (RL) to these massive models in large-scale distributed environments faces severe systemic bottlenecks, primarily due to the resource conflict between high-fidelity physical simulation and the intensive VRAM/bandwidth demands of deep learning. This conflict often leaves overall throughput constrained by execution-phase inefficiencies. To address these challenges, we propose D-VLA, a high-concurrency, low-latency distributed RL framework for large-scale embodied foundation models. D-VLA introduces "Plane Decoupling," physically isolating high-frequency training data from low-frequency weight control to eliminate interference between simulation and optimization. We further design a four-thread asynchronous "Swimlane" pipeline, enabling full parallel overlap of sampling, inference, gradient computation, and parameter distribution. Additionally, a dual-pool VRAM management model and topology-aware replication resolve memory fragmentation and optimize communication efficiency. Experiments on benchmarks like LIBERO show that D-VLA significantly outperforms mainstream RL frameworks in throughput and sampling efficiency for billion-parameter VLA models. In trillion-parameter scalability tests, our framework maintains exceptional stability and linear speedup, providing a robust system for high-performance general-purpose embodied agents.

Zipeng Liu, Likang Wu, Ming He et al.

Significant efforts have been dedicated to integrating the powerful Large Language Models (LLMs) with diverse modalities, particularly focusing on the fusion of language, vision and audio data. However, the graph-structured data, which is inherently rich in structural and domain-specific knowledge, has not yet been gracefully adapted to LLMs. Existing methods either describe the graph with raw text, suffering the loss of graph structural information, or feed Graph Neural Network (GNN) embeddings into LLMs at the cost of losing explainable prompt semantics. To bridge this gap, we introduce an end-to-end modality-aligning framework for LLM-graph alignment: Dual-Residual Vector Quantized-Variational AutoEncoder, namely Dr.E. Our approach is purposefully designed to facilitate token-level alignment with LLMs, enabling an effective translation of the intrinsic `language' of graphs into comprehensible natural language. We also manage to enhance LLMs' more robust structural understanding of graphs by incorporating multiple views of the central nodes based on their surrounding nodes at various distances. Our experimental evaluations on standard graph tasks demonstrate competitive performance against other state-of-the-art (SOTA) approaches. Additionally, our framework ensures certain visual interpretability, efficiency, and robustness, marking the promising successful endeavor to achieve token-level alignment between LLMs and GNNs. Our code is available at: https://github.com/Timothy914/Dr.E.

Wen Huang, Haoran Sun, Yongjian Guo et al.

World Action Models (WAMs) are an emerging family of policies that tie robot action generation to future-observation modeling. In this work, we focus on the joint video--action modeling paradigm, where actions and imagined future observations are co-generated along a shared denoising or flow trajectory, so that perception, prediction, and control are coupled within one generative process. Existing WAMs typically realize this paradigm with a Mixture-of-Transformers (MoT), where video and action tokens interact through shared self-attention. This architecture can in principle assign a separate timestep $t_f$ to each predicted latent frame, yet current systems collapse this degree of freedom onto a single shared scalar $t$. Under the noise-as-masking view of Diffusion Forcing, this shared schedule imposes the unjustified prior that every predicted latent is equally reliable for action generation. We instead view the per-latent schedule as a \emph{learnable information-gating policy}: by changing a latent frame's noise level, the policy modulates the reliability of its Key/Value contribution to the action tokens. We propose \textbf{NoiseGate}, which combines independent per-latent timestep sampling during backbone training, a lightweight Gating Policy Network that emits per-latent time increments during denoising, and task-reward optimization that trains the schedule policy without hand-crafted shape priors. Built on a joint video--action MoT backbone, NoiseGate delivers consistent gains on diverse RoboTwin random-scene manipulation tasks.

Jiaxuan Gao, Yongjian Guo, Zhong Guan et al.

The integration of Vision-Language-Action (VLA) models with World Models has gained increasing attention. One representative approach treats learned World Models as generative simulators, enabling policy optimization entirely within "imagination." However, when deployed as simulators for specific environments such as the LIBERO benchmark, existing World Models often suffer from poor generalization and long-horizon error accumulation. During closed-loop rollouts, these models are highly sensitive to initial-state perturbations; minor changes in color, illumination, and other visual factors can trigger cascading hallucinations, leading to severe blurriness or overexposure. Moreover, long-horizon error accumulation further degrades the quality and fidelity of predicted future states. These issues limit the reliability of World Models as simulators. To mitigate these problems, we propose Sword, a robust World Model framework. Our method introduces Structure-Guided Style Augmentation to disentangle the visual textures of interactive environments from task-relevant dynamics, thereby improving generalization. We further propose Dynamic Latent Bootstrapping, which maintains consistency between training and inference while keeping memory consumption low. Extensive experiments on the LIBERO benchmark show that our method significantly outperforms the baseline WoVR in terms of generalization, generation quality, robustness, fidelity, and the success rate of reinforcement-learning post-training for VLA models.

Zhong Guan, Likang Wu, Hongke Zhao et al.

Many existing studies have achieved significant improvements in the reasoning capabilities of large language models (LLMs) through reinforcement learning with verifiable rewards (RLVR), while the enhancement of reasoning abilities in small language models (SLMs) has not yet been sufficiently explored. Combining distilled data from larger models with RLVR on small models themselves is a natural approach, but it still faces various challenges and issues. Therefore, we propose \textit{\underline{R}}ecall-\textit{\underline{E}}xtend \textit{\underline{D}}ynamics(RED): Enhancing Small Language Models through Controlled Exploration and Refined Offline Integration. In this paper, we explore the perspective of varying exploration spaces, balancing offline distillation with online reinforcement learning. Simultaneously, we specifically design and optimize for the insertion problem within offline data. By monitoring the ratio of entropy changes in the model concerning offline and online data, we regulate the weight of offline-SFT, thereby addressing the issues of insufficient exploration space in small models and the redundancy and complexity during the distillation process. Furthermore, to tackle the distribution discrepancies between offline data and the current policy, we design a sample-accuracy-based policy shift mechanism that dynamically chooses between imitating offline distilled data and learning from its own policy.

Junlang Huang, Hao Chen, Zhong Guan

This paper proposes a neural framework for power and timing prediction of multi-stage data path, distinguishing itself from traditional lib-based analytical methods dependent on driver characterization and load simplifications. To the best of our knowledge, this is the first language-based, netlist-aware neural network designed explicitly for standard cells. Our approach employs two pre-trained neural models of waveform prediction and delay estimation that directly infer transient waveforms and propagation delays from SPICE netlists, conditioned on critical physical parameters such as load capacitance, input slew, and gate size. This method accurately captures both intrinsic and coupling-induced delay effects without requiring simplification or interpolation. For multi-stage timing prediction, we implement a recursive propagation strategy where predicted waveforms from each stage feed into subsequent stages, cumulatively capturing delays across the logic chain. This approach ensures precise timing alignment and complete waveform visibility throughout complex signal pathways. The waveform prediction utilizes a hybrid CNN-Transformer architecture with netlist-aware node-level encoding, addressing traditional Transformers' fixed input dimensionality constraints. Additionally, specialized subnetworks separately handle primary delay estimation and crosstalk correction. Experimental results demonstrate SPICE-level accuracy, consistently achieving RMSE below 0.0098 across diverse industrial circuits. The proposed framework provides a scalable, structurally adaptable neural alternative to conventional power and timing engines, demonstrating high fidelity to physical circuit behaviors.

Junlang Huang, Hao Chen, Li Luo et al.

Simulation of high-order nonlinear system requires extensive computational resources, especially in modern VLSI backend design where bifurcation-induced instability and chaos-like transient behaviors pose challenges. We present S-Crescendo - a nested transformer weaving framework that synergizes S-domain with neural operators for scalable time-domain prediction in high-order nonlinear networks, alleviating the computational bottlenecks of conventional solvers via Newton-Raphson method. By leveraging the partial-fraction decomposition of an n-th order transfer function into first-order modal terms with repeated poles and residues, our method bypasses the conventional Jacobian matrix-based iterations and efficiently reduces computational complexity from cubic $O(n^3)$ to linear $O(n)$.The proposed architecture seamlessly integrates an S-domain encoder with an attention-based correction operator to simultaneously isolate dominant response and adaptively capture higher-order non-linearities. Validated on order-1 to order-10 networks, our method achieves up to 0.99 test-set ($R^2$) accuracy against HSPICE golden waveforms and accelerates simulation by up to 18(X), providing a scalable, physics-aware framework for high-dimensional nonlinear modeling.

Junlang Huang, Hao Chen, Li Luo et al.

This paper presents a hybrid model combining Transformer and CNN for predicting the current waveform in signal lines. Unlike traditional approaches such as current source models, driver linear representations, waveform functional fitting, or equivalent load capacitance methods, our model does not rely on fixed simplified models of standard-cell drivers or RC loads. Instead, it replaces the complex Newton iteration process used in traditional SPICE simulations, leveraging the powerful sequence modeling capabilities of the Transformer framework to directly predict current responses without iterative solving steps. The hybrid architecture effectively integrates the global feature-capturing ability of Transformers with the local feature extraction advantages of CNNs, significantly improving the accuracy of current waveform predictions. Experimental results demonstrate that, compared to traditional SPICE simulations, the proposed algorithm achieves an error of only 0.0098. These results highlight the algorithm's superior capabilities in predicting signal line current waveforms, timing analysis, and power evaluation, making it suitable for a wide range of technology nodes, from 40nm to 3nm.

Zhong Guan, Hongke Zhao, Likang Wu et al.

Recently, large language models (LLMs) have been widely researched in the field of graph machine learning due to their outstanding abilities in language comprehension and learning. However, the significant gap between natural language tasks and topological structure modeling poses a nonnegligible challenge. Specifically, since natural language descriptions are not sufficient for LLMs to understand and process graph-structured data, fine-tuned LLMs perform even worse than some traditional GNN models on graph tasks, lacking inherent modeling capabilities for graph structures. Existing research overly emphasizes LLMs' understanding of semantic information captured by external models, while inadequately exploring graph topological structure modeling, thereby overlooking the genuine capabilities that LLMs lack. Consequently, in this paper, we introduce a new framework, LangTopo, which aligns graph structure modeling with natural language understanding at the token level. LangTopo quantifies the graph structure modeling capabilities of GNNs and LLMs by constructing a codebook for the graph modality and performs consistency maximization. This process aligns the text description of LLM with the topological modeling of GNN, allowing LLM to learn the ability of GNN to capture graph structures, enabling LLM to handle graph-structured data independently. We demonstrate the effectiveness of our proposed method on multiple datasets.

Zhong Guan, Likang Wu, Hongke Zhao et al.

Large Language Models (LLMs) are increasingly prominent in the recommendation systems domain. Existing studies usually utilize in-context learning or supervised fine-tuning on task-specific data to align LLMs into recommendations. However, the substantial bias in semantic spaces between language processing tasks and recommendation tasks poses a nonnegligible challenge. Specifically, without the adequate capturing ability of collaborative information, existing modeling paradigms struggle to capture behavior patterns within community groups, leading to LLMs' ineffectiveness in discerning implicit interaction semantic in recommendation scenarios. To address this, we consider enhancing the learning capability of language model-driven recommendation models for structured data, specifically by utilizing interaction graphs rich in collaborative semantics. We propose a Graph-Aware Learning for Language Model-Driven Recommendations (GAL-Rec). GAL-Rec enhances the understanding of user-item collaborative semantics by imitating the intent of Graph Neural Networks (GNNs) to aggregate multi-hop information, thereby fully exploiting the substantial learning capacity of LLMs to independently address the complex graphs in the recommendation system. Sufficient experimental results on three real-world datasets demonstrate that GAL-Rec significantly enhances the comprehension of collaborative semantics, and improves recommendation performance.