arXiv subjects

Song Chen

Publications and source records attributed to Song Chen.

At least 19 recordsLinked to original sources

arXiv2026

Near-Unity Excitation and Radiative Efficiencies in Electroluminescence Without External Carrier Injection

Electroluminescence occurring without external charge injection is typically characterized by weak emission and excessive driving voltage, due to low excitation and radiative recombination efficiencies. Here, we demonstrate non-injecting electroluminescence (NI-EL) that challenges this conventional perception. To achieve this, we introduce an operational paradigm that leverages remote, state-abundant charge reservoirs, which elevates the excitation efficiency close to unity - a greater-than-20-fold improvement over the benchmark. This strategy is augmented by quantum dots (QDs) with nonmonotonically graded shells, raising the high-field radiative efficiency by approximately 7-8-fold. The resultant RGB NI-EL devices uniformly exhibit bright and efficient pulsed emissions, with key metrics including: a turn-on threshold of 3.7 Vrms for red; a luminance of 291,628 nits and a power efficiency of 302.6 lm/W for green, with light-outcoupling enhancement; and, for blue, the superior emitter stability of the first QD-based blue NI-EL over its light-emitting diode counterpart. The minimal dielectric loss, sub-100 ns response time, and external EL efficiency of up to 45.4% further reinforce the mechanism-performance causality. These results overcome the inherent mechanistic limitations of NI-EL and establish performance that rivals or surpasses injection-type EL, including AC- or DC-driven variants, positioning NI-EL as a promising platform for high-performance pulsed light sources.

physics.optics↗

arXiv2026

A Continuous-Time Analysis of Smoothed Matrix-Polar Spectral Gradient Flows for Muon-Type Optimization

This paper studies smoothed matrix-polar spectral gradient flows for unconstrained matrix-valued optimization.The canonical polar-factor map loses smoothness at rank-deficient matrices and becomes ill-conditioned as singular values approach zero, creating analytical difficulties.We therefore introduce a spectral feedback law generated by a smooth spectral potential and establish the regularity, monotonicity, boundedness, and dissipation properties of the feedback.Based on this feedback law, we propose a smoothed spectral gradient flow and prove well-posedness and global convergence of the flow.We derive convergence-rate results for the spectral gradient flow in nonconvex, convex, and Polyak--Lojasiewicz (PL) settings and analyze the Lyapunov structure and convergence of a momentum-augmented system under the same spectral feedback law. Furthermore, we provide a local descent-rate comparison between the smoothed spectral-gradient direction and the standard Frobenius-gradient direction using a general Hessian-based quadratic model. This analysis yields a verifiable normalized descent-rate advantage condition, showing that the local benefit of the spectral direction depends on both first-order alignment with the gradient matrix and the directional curvature induced by the Hessian.

math.OC↗

arXiv2026

DART: Decoded Attention over Recurrent States for Efficient Long-Context Sequence Modeling

Modern language models are built primarily from Transformers, recurrent models, and their hybrid architectures. Transformers rely on token-level attention memories, while recurrent models such as state space models (SSMs) and linear attention maintain compact recurrent states. These architectures are typically instantiated separately or interleaved at the layer level, leaving open whether a shared memory representation can support both recurrent compression and attention-style retrieval. We study this question through the state space duality (SSD) view of Mamba-2, where the SSM state can be interpreted as a compressed associative key--value (KV) cache. We observe that Mamba-2 decodes token-conditioned values from this state but does not decode token-conditioned keys. Based on this observation, we propose DART (Decoded Attention over Recurrent sTates), which retains the chunk state contributions produced by the Mamba-2 chunked scan as chunk state memories, decodes token-conditioned keys and values from these memories, and performs state-memory attention (SMA) over the resulting KV pairs. The retrieved output is then combined with the native Mamba-2 output through a gated residual connection. DART supports practical training by reusing the Mamba-2 chunked scan and implementing SMA as a FlashAttention-style computation. Our analysis and experiments show that DART substantially reduces the length-dependent inference cache compared with a matched attention baseline (e.g., $75\%$ savings when the chunk size is $S=256$ and the state size is $N=128$). Compared with Mamba-2, DART substantially improves associative recall and retrieval while preserving general language-modeling quality.

cs.LG↗

arXiv2026

DSSMs: State Space Models with Explicit Memory via Delay Differential Equations

State Space Models (SSMs) have emerged as a powerful paradigm for efficient long-sequence modeling, offering parallel training and fast linear-time recurrent inference. However, like other recurrent architectures, SSMs must compress an unbounded history into a fixed-size state, which limits context retention and makes precise retrieval over long-range context inherently difficult. To overcome this limitation, we propose Delay State Space Models (DSSMs), a delay differential equation (DDE)-inspired extension of diagonal SSMs that augments discrete SSM recurrences with explicit delayed-state feedback. Making explicit delayed feedback practical requires new stability parameterization, history management, and FFT-training tools. We address these challenges with a practical discretization and parameterization grounded in a simple delay-independent stability condition. To bypass direct time-domain kernel construction, we derive the DSSM transfer function and compute kernels in the frequency domain, using a kernel contour shift to suppress aliasing and recover accurate FFT training. Empirically, DSSMs substantially improve targeted delayed-retrieval tasks while outperforming S4D on most standard sequence metrics and remaining close on the others.

cs.LG↗

arXiv2026

A Backstepping Framework for Unconstrained Accelerated Optimization Algorithms

This paper introduces a control-theoretic perspective on unconstrained optimization algorithms using the backstepping methods. We model the optimization process as an augmented strict-feedback system given by $\dot{x}_1 = x_2$, $\dot{x}_2 = u$, and $\dot{z} = q(x_1,z)$, with a regulated output $y = \nabla f(x_1)$. This formulation recasts the development of unconstrained optimization algorithms as a feedback control problem, where the goal is to design the input $u$ to ensure $y(t) \to 0$. By employing backstepping, we recursively synthesize the actual feedback law $u$ after initially selecting a virtual control for $x_1$. For convex objective functions, we develop a general synthesis framework for augmented strict-feedback systems and specialize it to the standard strict-feedback case. This unified framework successfully recovers the constant-parameter Nesterov flow and the proportional-integral-derivative (PID) accelerated optimizer as direct corollaries. We further establish that, given a fixed virtual control, the universal second-step law is inverse optimal with respect to an induced outer-tracking problem. This reveals that the optimality of the control law is conditionally dependent on the target manifold prescribed by the virtual control, rather than holding globally across all possible backstepping designs. Finally, we formulate a formal optimal-backstepping theorem that elevates this optimality principle to the virtual-control stage by solving a reduced Hamilton--Jacobi--Bellman problem. These contributions collectively yield a robust and general backstepping-driven paradigm for the analysis and design of continuous-time unconstrained optimization algorithms.

math.OC↗

arXiv2026

Swarm-Inspired Generation of Collective Behaviors in Graph Dynamical Systems

Collective behavior arises when locally interacting units produce coordinated global organization, from synchronization in dynamical systems to task-relevant information flow on graphs. The central challenge is not only to explain how collective behavior emerges, but to design local interaction rules that can produce desired global organization and generalize across graphs, dynamics and tasks.To address this challenge, we introduce the Swarm-Inspired Emergent Synchronizer (SIES), a graph-dynamical framework that learns generalizable local-interaction laws for controllable collective organization. Each node is an agent-like dynamical unit with a state and task cue, and signed source-target-conditioned attention acts as an adaptive coupling term inside an explicit evolution model. Therefore, SIES combines an explicit dynamical engine with local agent intelligence, similar to biological swarms. For synchronization control, SIES learns a generalizable coupling operator that produces prescribed synchronization patterns for CDSs across untrained network scales, target phase relations, and intrinsic node dynamics without retraining. The learned operator also reaches gait-related modes faster than three oscillator baselines and generalizes synchronization-driven locomotion to simulated multi-legged robots of different scales and a physical hexapod after leg disablement. For graph representation learning, SIES applies the same signed interaction principle to message passing and achieves the highest performance among the compared methods on heterophilous node-classification benchmarks. Together, these results position SIES as a generalizable and learnable graph-dynamical interaction framework with promise for synchronization control, adaptive robot coordination, and heterophilous graph representation learning.

cs.LG↗

arXiv2026

VICX: Generalizable Robot Manipulation via Video Generation and In-Context Operator Network

Generalizable robot manipulation requires not only task-level reasoning over unseen scenes, but also reliable grounding of visual plans into embodiment-specific execution. To bridge this gap, we propose VICX (Video generation and In-Context eXecution), a decoupled closed-loop manipulation framework. In VICX, a frozen video generation model produces vision-language-conditioned high-level visual plans, while a Video-to-Trajectory In-Context Operator Network (V2T-ICON) serves as the task-agnostic interface that grounds these plans into executable robot-state trajectories. To improve execution generalization, V2T-ICON operates on segmentation-extracted arm-only frame observations and uses retrieved image-state pairs as in-context prompts, allowing a robust and generalizable visual-to-state mapping at inference time without parameter updates. Experiments on Meta-World show that VICX supports cross-task generalization, closed-loop self-correction, and cross-embodiment transfer, demonstrating dual generalization across both task semantics and robot execution. The project webpage can be found here: https://scaling-group.github.io/vicx/.

cs.RO↗

arXiv2026

BitsMoE: Efficient Spectral Energy-Guided Bit Allocation for MoE LLM Quantization

Mixture-of-Experts (MoE) large language models reduce per-token computation through sparse expert activation, but their deployment remains memory-intensive because all expert weights must be kept resident in memory. Existing MoE compression methods struggle in the ultra-low-bit regime: pruning irreversibly removes model capacity, while coarse-grained quantization fails to allocate bits according to heterogeneous expert and weight-direction importance. We propose BitsMoE, a spectral-energy-guided bit-allocation framework for MoE LLM quantization. BitsMoE decomposes each MoE layer by SVD into a shared basis and expert-specific spectral factors, retaining the shared basis without quantization to preserve common cross-expert structure and using the expert-specific factors as fine-grained quantization units. To determine the bit-width of each unit, BitsMoE formulates spectrum-wise mixed-precision quantization as an activation-aware reconstruction surrogate and solves an integer linear program that minimizes estimated reconstruction loss under a fixed bit budget. Experiments across multiple MoE LLMs show that BitsMoE substantially reduces downstream task accuracy degradation in ultra-low-bit regimes. Under 2-bit quantization on Qwen3-30B-A3B-Base, BitsMoE accelerates quantization by 12.3$\times$, improves average accuracy by 27.83 percentage points, and increases decoding speed by 1.76$\times$ over GPTQ. Our model and code are publicly available at https://github.com/zjiayu064/BitsMoE.

cs.LG↗

arXiv2026

DORA: Dynamic Online Reinforcement Agent for Token Merging in Vision Transformers

Vision Transformers (ViTs) incur significant computational overhead due to the quadratic complexity of self-attention relative to the token sequence length. While existing token reduction methods mitigate this issue, they predominantly rely on fixed heuristic metrics, predefined ratios, or static offline masks, which lack the adaptability to capture input-dependent redundancy during inference. In this paper, we propose DORA (Dynamic Online Reinforcement Agent), the first reinforcement learning (RL)-driven online inference framework for dynamic token merging in ViTs. We formulate the merging process as a sequential Markov Decision Process (MDP), where a lightweight RL agent determines the merging strategy for each Transformer block based on the current feature state and layer-specific context. To balance computational efficiency and feature fidelity, the agent is optimized via a dense reward function incorporating a non-linear distillation-based penalty. We implement an asymmetric Actor-Critic architecture that utilizes a high-capacity Critic for stable offline training while retaining a minimal Actor head for low-computation online inference. Evaluations across multiple ViT scales (Tiny to Large) demonstrate that DORA improves the accuracy-efficiency Pareto front compared to current baselines. Under strict negligible accuracy-drop constraints (<= 0.05%), DORA achieves up to a 12.66% token merging rate, and delivers up to a 569.7% relative improvement over the most efficient baseline. On ImageNet-1K, under aligned accuracy constraints, DORA achieves up to a 76% relative improvement in computational savings compared to state-of-the-art methods. Furthermore, on out-of-distribution (OOD) benchmarks such as ImageNet-A and ImageNet-C, DORA attains a relative efficiency advantage of over 430%.

cs.CV↗

arXiv2026

Rotated Robustness: A Training-Free Defense against Bit-Flip Attacks on Large Language Models

Hardware faults, specifically bit-flips in quantized weights, pose a severe reliability threat to Large Language Models (LLMs), often triggering catastrophic model collapses. We demonstrate that this vulnerability fundamentally stems from the spatial alignment between sensitive weight bits and extreme activation outliers, which causes a single hardware fault to be massively amplified. To address this, we propose Rotated Robustness (RoR), a training-free defense utilizing orthogonal Householder transformations. By applying an orthogonal rotation to the activation space, RoR geometrically smooths extreme outliers across all feature dimensions. This mechanism effectively breaks the alignment between outliers and vulnerable weights, mathematically guaranteeing original model accuracy. Extensive empirical evaluations across Llama-2/3, OPT, and Qwen families demonstrate the superior reliability of our approach. Under random bit-flip attacks, RoR reduces the stochastic collapse rate from 3.15\% to 0.00\% on Qwen2.5-7B. Furthermore, under severe targeted attacks with 50 Progressive Bit Search flips, RoR sustains robust reasoning on Llama-2-7B, maintaining a 43.9\% MMLU accuracy that nearly matches its 45.2\% unattacked accuracy, while competing defenses collapse to random guessing. Most notably, against the Single-Point Fault Attack (SPFA) -- the most aggressive targeted threat -- RoR exponentially inflates the attack complexity from a few bits to over 17,000 precise bit-flips. With a negligible storage overhead of 0.31\% and a minimal inference latency increase of 9.1\% on Llama-2-7B, RoR achieves true lossless robustness, providing a practical and highly reliable defense for LLM deployment.

cs.CR↗

arXiv2026

An FPGA Implementation of Displacement Vector Search for Intra Pattern Copy in JPEG XS

Recently, progress has been made on the Intra Pattern Copy (IPC) tool for JPEG XS, an image compression standard designed for low-latency and low-complexity coding. IPC performs wavelet-domain intra compensation predictions to reduce spatial redundancy in screen content. A key module of IPC is the displacement vector (DV) search, which aims to solve the optimal prediction reference offset. However, the DV search process is computationally intensive, posing challenges for practical hardware deployment. In this paper, we propose an efficient pipelined FPGA architecture design for the DV search module to promote the practical deployment of IPC. Optimized memory organization, which leverages the IPC computational characteristics and data inherent reuse patterns, is further introduced to enhance the performance. Experimental results show that our proposed architecture achieves a throughput of 38.3 Mpixels/s with a power consumption of 277 mW, demonstrating its feasibility for practical hardware implementation in IPC and other predictive coding tools, and providing a promising foundation for ASIC deployment.

cs.AR↗

arXiv2026

SecAgent: Efficient Mobile GUI Agent with Semantic Context

Mobile Graphical User Interface (GUI) agents powered by multimodal large language models have demonstrated promising capabilities in automating complex smartphone tasks. However, existing approaches face two critical limitations: the scarcity of high-quality multilingual datasets, particularly for non-English ecosystems, and inefficient history representation methods. To address these challenges, we present SecAgent, an efficient mobile GUI agent at 3B scale. We first construct a human-verified Chinese mobile GUI dataset with 18k grounding samples and 121k navigation steps across 44 applications, along with a Chinese navigation benchmark featuring multi-choice action annotations. Building upon this dataset, we propose a semantic context mechanism that distills history screenshots and actions into concise, natural language summaries, significantly reducing computational costs while preserving task-relevant information. Through supervised and reinforcement fine-tuning, SecAgent outperforms similar-scale baselines and achieves performance comparable to 7B-8B models on our and public navigation benchmarks. Our dataset is available at https://huggingface.co/datasets/alibabagroup/CMGUI.

cs.CV↗

arXiv2026

Distributed physics-informed neural networks via domain decomposition for fast flow reconstruction

Physics-Informed Neural Networks (PINNs) offer a powerful paradigm for flow reconstruction, seamlessly integrating sparse velocity measurements with the governing Navier-Stokes equations to recover complete velocity and latent pressure fields. However, scaling such models to large spatiotemporal domains is hindered by computational bottlenecks and optimization instabilities. In this work, we propose a robust distributed PINNs framework designed for efficient flow reconstruction via spatiotemporal domain decomposition. A critical challenge in such distributed solvers is pressure indeterminacy, where independent sub-networks drift into inconsistent local pressure baselines. We address this issue through a reference anchor normalization strategy coupled with decoupled asymmetric weighting. By enforcing a unidirectional information flow from designated master ranks where the anchor point lies to neighboring ranks, our approach eliminates gauge freedom and guarantees global pressure uniqueness while preserving temporal continuity. Furthermore, to mitigate the Python interpreter overhead associated with computing high-order physics residuals, we implement a high-performance training pipeline accelerated by CUDA graphs and JIT compilation. Extensive validation on complex flow benchmarks demonstrates that our method achieves near-linear strong scaling and high-fidelity reconstruction, establishing a scalable and physically rigorous pathway for flow reconstruction and understanding of complex hydrodynamics.

cs.LG↗

arXiv2025

Synergy and Competition of Dual Chirality in the Chirality-Induced Spin Selectivity of Supramolecular Helices

Recent progress in constructing supramolecular assemblies with hierarchical chirality offers new opportunities to investigate the chirality-induced spin selectivity (CISS) effect and its potential applications. In this work, we systematically examine the CISS effect in such multichiral systems by designing a class of multilayer helical architectures constructed of stacked and interfaced individual helical rings, each possessing well-defined local chirality. Through controlled interlayer twisting, a global helical handedness is further imposed, forming a multichiral tubular helix. Theoretical calculations reveal that these two distinct chiral hierarchies lead to several unprecedented CISS phenomena, such as enhanced spin polarization arising from cooperative dual chirality, along with the simultaneous emergence of transverse and longitudinal CISS signals. Moreover, interlayer torsional competition modulates the system's response to external fields. The dual-chiral geometry breaks the conventional symmetry of single helices, inducing an anomalous angular phase shift in magnetoresistance. Furthermore, Floquet analysis reveals that the interplay between local and global chirality enables controlled spin polarization switching under circularly polarized light. These findings provide a basic theoretical framework for studying the CISS in multichiral superstructures and establish design principles for coupled optical, magnetic, and spin manipulations, thereby facilitating the development of multichiral spintronic devices.

cond-mat.mes-hall↗

arXiv2025

SASQ: Static Activation Scaling for Quantization-Aware Training in Large Language Models

Large language models (LLMs) excel at natural language tasks but face deployment challenges due to their growing size outpacing GPU memory advancements. Model quantization mitigates this issue by lowering weight and activation precision, but existing solutions face fundamental trade-offs: dynamic quantization incurs high computational overhead and poses deployment challenges on edge devices, while static quantization sacrifices accuracy. Existing approaches of quantization-aware training (QAT) further suffer from weight training costs. We propose SASQ: a lightweight QAT framework specifically tailored for activation quantization factors. SASQ exclusively optimizes only the quantization factors (without changing pre-trained weights), enabling static inference with high accuracy while maintaining deployment efficiency. SASQ adaptively truncates some outliers, thereby reducing the difficulty of quantization while preserving the distributional characteristics of the activations. SASQ not only surpasses existing SOTA quantization schemes but also outperforms the corresponding FP16 models. On LLaMA2-7B, it achieves 5.2% lower perplexity than QuaRot and 4.7% lower perplexity than the FP16 model on WikiText2.

cs.CL↗

arXiv2025

HQ-DM: Single Hadamard Transformation-Based Quantization-Aware Training for Low-Bit Diffusion Models

Diffusion models have demonstrated significant applications in the field of image generation. However, their high computational and memory costs pose challenges for deployment. Model quantization has emerged as a promising solution to reduce storage overhead and accelerate inference. Nevertheless, existing quantization methods for diffusion models struggle to mitigate outliers in activation matrices during inference, leading to substantial performance degradation under low-bit quantization scenarios. To address this, we propose HQ-DM, a novel Quantization-Aware Training framework that applies Single Hadamard Transformation to activation matrices. This approach effectively reduces activation outliers while preserving model performance under quantization. Compared to traditional Double Hadamard Transformation, our proposed scheme offers distinct advantages by seamlessly supporting INT convolution operations while preventing the amplification of weight outliers. For conditional generation on the ImageNet 256x256 dataset using the LDM-4 model, our W4A4 and W4A3 quantization schemes improve the Inception Score by 12.8% and 467.73%, respectively, over the existing state-of-the-art method.

cs.CV↗

arXiv2025

VeriGRAG: Enhancing LLM-Based Verilog Code Generation with Structure-Aware Soft Prompts

Large language models (LLMs) have demonstrated strong capabilities in generating Verilog code from natural language descriptions. However, Verilog code inherently encodes structural information of hardware circuits. Effectively leveraging this structural information to enhance the functional and syntactic correctness of LLM-generated Verilog code remains a significant challenge. To address this challenge, we propose VeriGRAG , a novel framework that extracts structural graph embeddings from Verilog code using graph neural networks (GNNs). A multimodal retriever then selects the graph embeddings most relevant to the given generation task, which are aligned with the code modality through the VeriFormer module to generate structure-aware soft prompts. Our experiments demonstrate that VeriGRAG substantially improves the correctness of Verilog code generation, achieving state-of-the-art or superior performance across both VerilogEval and RTLLM benchmarks.

cs.AR↗

arXiv2025

Accelerated Distributed Aggregative Optimization

This paper delves into the investigation of a distributed aggregative optimization problem within a network. In this scenario, each agent possesses its own local cost function, which relies not only on the local state variable but also on an aggregated function of state variables from all agents. To expedite the optimization process, we amalgamate the heavy ball and Nesterovs accelerated method with distributed aggregative gradient tracking, resulting in the proposal of two innovative algorithms, aimed at resolving the distributed aggregative optimization problem. Our analysis demonstrates that the proposed algorithms can converge to an optimal solution at a global linear convergence rate when the objective function is strongly convex with the Lipschitz-continuous gradient, and when the parameters (e.g., step size and momentum coefficients) are chosen within specific ranges. Additionally, we present several numerical experiments to verify the effectiveness, robustness and superiority of our proposed algorithms.

math.OC↗