arXiv subjects

Haowei Li

Publications and source records attributed to Haowei Li.

At least 19 recordsLinked to original sources

arXiv2026

EvoSCM: Scientific Belief Revision Through Causal Model Evolution and Experimentation

Scientific agents must learn not only how to reason, but also what to believe. However, existing LLM agents typically express scientific hypotheses in free-form text, leaving their beliefs implicit and difficult to test or revise. We introduce EvoSCM, which equips scientific agents with explicit structural causal models that evolve as new experimental evidence is collected. EvoSCM maintains a population of competing SCM hypotheses, each encoding a candidate causal explanation of the environment, and evolves them through a closed discovery loop. In each round, the agent abduces latent mechanisms from accumulated evidence, designs discriminative interventions, and commits to falsifiable predictions that it tests through experimentation. Discrepancies between prediction and observation are inductively distilled into correction rules that revise the causal structures and mechanisms of each hypothesis, and the agent then deductively validates the revised population against accumulated evidence and structural consistency to guide the next round. We evaluate EvoSCM on DiscoverPhysics, a benchmark requiring agents to uncover the hidden dynamics of noncanonical physical worlds through experimentation. EvoSCM consistently improves scientific discovery over baselines, yielding more accurate explanations and predictions while making more effective use of experimental interactions.

cs.AI↗

arXiv2026

Hierarchical Shared Memory-Aware Optimization for TRSM on GPU Platforms

Triangular Solve with Multiple Right-hand Sides (TRSM) is a fundamental BLAS Level-3 operation that underpins LU/Cholesky decomposition, sparse direct solvers, and matrix inversion. In the left-side lower-triangular case studied in this paper, efficient GPU implementation remains challenging because forward substitution introduces strict row-wise dependencies, and shared memory is too scarce to hold both operand matrices for wide data types such as double complex. This paper presents HSMA-TRSM, a hierarchical shared memory-aware optimization framework for left-side lower-triangular TRSM on NVIDIA A100, NVIDIA H800, and Hygon DCU Z100 accelerators. For the small-scale regime (m,n<=64), we design a pipelined compute-memory overlap mechanism through loop unrolling and instruction reordering, and propose a dual thread-group seven-stage pipeline strategy to address shared memory constraints for double complex types. For large-scale problems, we introduce a diagonal block decoupling optimization with an O(IB)shared-memory footprint for diagonal block inversion, enabling adaptive block size selection based on matrix scale and hardware characteristics. A compile-time configuration selection framework based on offline profiling and online lookup selects the optimal block size per platform with zero runtime overhead. Evaluated on NVIDIA A100, H800, and Hygon DCU Z100, HSMA-TRSM achieves peak speedups of 2.05xover cuBLAS and 2.06xover rocBLAS. The gains are strongest in shared-memory-constrained double-complex small cases and in large real-type cases where adaptive blocking improves GEMM-dominated updates, while mature vendor kernels leave less optimization headroom in some regimes.

cs.DC↗

arXiv2026

Kibble--Zurek Scaling in the Dicke Model at Mesoscopic Scales

The Dicke model is a paradigmatic setting for collective light-matter physics and the superradiant phase transition. Yet extracting the critical exponents is challenging at experimentally accessible mesoscopic sizes, due to the slow divergence of the correlation time under all-to-all coupling and a photon-loss-driven crossover to a distinct dissipative universality class. Here, we perform a large-$N$ analysis that identifies distinct coherent and dissipative fixed points for the closed and open Dicke models. We then develop a unified mesoscopic scaling framework that incorporates the leading irrelevant correction and, going beyond static and spectral probes, brings ramping dynamics under the same scaling description. It recovers the corresponding exponents, verifies Kibble-Zurek scaling, and clarifies how finite size, dissipation, and speed compete in the ramping dynamics. Our work thus establishes a unified framework for resolving static and dynamical critical scaling in closed and open quantum systems, with broader applicability to mesoscopic systems with long-range interactions.

cond-mat.quant-gas↗

arXiv2026

Tunable Memory Effect in Dissipative Strongly Correlated Quantum Systems

Strongly interacting quantum many-body systems subjected to non-Markovian dissipation pose a formidable challenge due to the interplay between strong correlation effects and memory effects. In this Letter, we develop a general theoretical framework to compute how a system observable responds to dissipation, which captures memory effects at short times and recovers the Markovian limit at longer times. Using this framework, we predict that, for a strongly correlated quantum critical state with critical exponent $\eta$, the short-time dynamics of a system observable always obeys a $t^{2\eta}$ scaling law. This emerges as a universal result from the interplay between strong correlation and memory effects, independent of the microscopic Hamiltonian of the system. We further reveal a crossover behavior of this scaling law to either $t^{2\eta-1}$ or linear-in-$t$ behavior beyond the memory time scale. We propose a concrete physical realization of a non-Markovian bath with tunable memory time using ultracold atoms, where our predictions can be straightforwardly verified in current experiments.

cond-mat.quant-gas↗

arXiv2026

One Scene, Two Depths: Probing Geometric Ambiguity in Monocular Foundation Models

A faithful 3D world representation should account for layered geometry, where a single camera ray may contain multiple visible and geometrically valid surfaces. Monocular depth estimation, however, reduces this structure to one scalar depth per pixel. Transparent scenes make this ambiguity measurable: the same ray can pass through foreground glass and observe the background, turning the supervised target into a convention of annotation, data, and training rather than a scene-intrinsic truth. A learned predictor exposes this convention as its depth-layer preference. We introduce MultiDepth-3k (MD-3k), a sparse two-layer ordinal benchmark for measuring depth-layer preference and multi-layer spatial relationship accuracy (ML-SRA). On MD-3k, leading depth foundation models exhibit diverse layer preferences under standard RGB input, showing that the same layered geometry can be resolved differently across models. We further find that Laplacian Visual Prompting (LVP), a training-free spectral input transformation, can substantially change the reported layer for certain frozen models. The strongest RGB/LVP pair, DAv2-L, reaches 75.5% ML-SRA. These results suggest that depth foundation models may express complementary geometric hypotheses that standard RGB inference leaves unexpressed. We invite the community to rethink depth supervision and evaluation through an ambiguity-aware lens, where multiple valid 3D interpretations are treated as geometric structure to be measured, preserved, and expressed.

cs.CV↗

arXiv2026

SPADE-Bench: Evaluating Spontaneous Strategic Deception in Agents via Plan-Action Divergence

As LLM-based agents expand their operational scope, reliability becomes a prerequisite for real-world deployment. However, in practical applications, human users cannot monitor every immediate behavior; instead, the execution process often remains a black box, leaving users dependent solely on the agent's self-reported updates. This opacity creates a critical risk: agents may present observer-facing reports that diverge from their executed actions, rendering the system uncontrollable, especially in high-stakes autonomous scenarios. We term such self-reported plan-action divergence as agent deception. To assess this, we introduce SPADE-Bench, a benchmark designed to evaluate spontaneous plan-action divergence. Unlike prior deception benchmarks, SPADE-Bench simultaneously integrates actual tool execution and controlled pressure scenarios. This design ensures ecological validity and rigorously distinguishes strategic deception from mere hallucination through controlled plan-action comparisons under pressure. Experiments across mainstream models confirm that agent deception is a genuine and pressing issue in tool-use contexts. By providing a comprehensive and robust evaluation framework, SPADE-Bench fills a critical gap in agent safety, facilitating the community's progress toward building trustworthy and controllable autonomous systems.

cs.CL↗

arXiv2026

LoPRo: Enhancing Low-Rank Quantization via Permuted Block-Wise Rotation

Post-training quantization (PTQ) enables effective model compression while preserving relatively high accuracy. Current weight-only PTQ methods primarily focus on the challenging sub-3-bit regime, where approaches often suffer significant accuracy degradation, typically requiring fine-tuning to achieve competitive performance. In this work, we revisit the fundamental characteristics of weight quantization and analyze the challenges in quantizing the residual matrix under low-rank approximation. We propose LoPRo, a novel fine-tuning-free PTQ algorithm that enhances residual matrix quantization by applying block-wise permutation and Walsh-Hadamard transformations to rotate columns of similar importance, while explicitly preserving the quantization accuracy of the most salient column blocks. Furthermore, we introduce a mixed-precision fast low-rank decomposition based on rank-1 sketch (R1SVD) to further minimize quantization costs. Experiments demonstrate that LoPRo outperforms existing fine-tuning-free PTQ methods at both 2-bit and 3-bit quantization, achieving accuracy comparable to fine-tuning baselines. Specifically, LoPRo achieves state-of-the-art quantization accuracy on LLaMA-2 and LLaMA-3 series models while delivering up to a 4$\times$ speedup. In the MoE model Mixtral-8x7B, LoPRo completes quantization within 2.5 hours, simultaneously reducing perplexity by 0.4$\downarrow$ and improving accuracy by 8\%$\uparrow$. Moreover, compared to other low-rank quantization methods, LoPRo achieves superior accuracy with a significantly lower rank, while maintaining high inference efficiency and minimal additional latency.

cs.LG↗

arXiv2025

Humanoid-inspired Causal Representation Learning for Domain Generalization

This paper proposes the Humanoid-inspired Structural Causal Model (HSCM), a novel causal framework inspired by human intelligence, designed to overcome the limitations of conventional domain generalization models. Unlike approaches that rely on statistics to capture data-label dependencies and learn distortion-invariant representations, HSCM replicates the hierarchical processing and multi-level learning of human vision systems, focusing on modeling fine-grained causal mechanisms. By disentangling and reweighting key image attributes such as color, texture, and shape, HSCM enhances generalization across diverse domains, ensuring robust performance and interpretability. Leveraging the flexibility and adaptability of human intelligence, our approach enables more effective transfer and learning in dynamic, complex environments. Through both theoretical and empirical evaluations, we demonstrate that HSCM outperforms existing domain generalization models, providing a more principled method for capturing causal relationships and improving model robustness. The code is available at https://github.com/lambett/HSCM.

cs.AI↗

arXiv2025

End2Race: Efficient End-to-End Imitation Learning for Real-Time F1Tenth Racing

F1Tenth is a widely adopted reduced-scale platform for developing and testing autonomous racing algorithms, hosting annual competitions worldwide. With high operating speeds, dynamic environments, and head-to-head interactions, autonomous racing requires algorithms that diverge from those in classical autonomous driving. Training such algorithms is particularly challenging: the need for rapid decision-making at high speeds severely limits model capacity. To address this, we propose End2Race, a novel end-to-end imitation learning algorithm designed for head-to-head autonomous racing. End2Race leverages a Gated Recurrent Unit (GRU) architecture to capture continuous temporal dependencies, enabling both short-term responsiveness and long-term strategic planning. We also adopt a sigmoid-based normalization function that transforms raw LiDAR scans into spatial pressure tokens, facilitating effective model training and convergence. The algorithm is extremely efficient, achieving an inference time of less than 0.5 milliseconds on a consumer-class GPU. Experiments in the F1Tenth simulator demonstrate that End2Race achieves a 94.2% safety rate across 2,400 overtaking scenarios, each with an 8-second time limit, and successfully completes overtakes in 59.2% of cases. This surpasses previous methods and establishes ours as a leading solution for the F1Tenth racing testbed. Code is available at https://github.com/michigan-traffic-lab/End2Race.

cs.RO↗

arXiv2025

Cavity-Mediated Gas-Liquid Transition

We study the gas-liquid transition in a binary Bose-Einstein condensate, where the two Zeeman-shifted hyperfine spin components are coupled by cavity-assisted Raman processes. Below a critical Zeeman field, the cavity becomes superradiant for an infinitesimally small pumping strength, where the enhanced superradiance is facilitated by the simultaneous formation of quantum droplet, a self-bound liquid phase stabilized by quantum fluctuations. Above the critical Zeeman field, the gas-liquid transition only takes place at a finite pumping strength after the system becomes superradiant. As the back action of the gas-liquid transition, the superradiant cavity field undergoes an abrupt jump at the first-order transition point. Furthermore, as a result of the fixed density ratio of the quantum droplet, the cavity field exhibits a linear scaling with the pumping strength in the liquid phase. These features serve as prominent signals for the cavity-mediated gas-liquid transition and coexistence, which derive from the interplay of Zeeman field, cavity-assisted spin mixing, and quantum fluctuations.

cond-mat.quant-gas↗

arXiv2025

Bridging the classical and quantum regimes in a dissipative Ising chain

We study the long-time dynamics of a dissipative Ising chain with varying quantum correlation. Invoking an ensemble-average formalism, and assuming spatial translation symmetry, we show that the dynamics can be described by a Lindblad master equation with an interpolated coherent Hamiltonian. In the classical limit, the interpolation Hamiltonian leads to a set of nonlinear equations of motion, where limit cycles can emerge in the long-time dynamics. In the quantum limit, by contrast, the system approaches a ferromagnetic steady state at long times. In between the two extremes, the discrete spatial translation symmetry can be spontaneously broken, as an antiferromagnetic steady state emerges, bridging the classical and quantum regimes. In particular, we illustrate how the classical limit-cycle behavior gradually disappears with the increase of quantum correlation. Since our model in the two extremes respectively applies to a dissipative Rydberg gas in the high- and zero-temperature limits, we expect it to provide a qualitatively correct description of dissipative Rydberg gases at interim temperatures, and shed light on the fate of limit cycles in a quantum open system.

quant-ph↗

arXiv2025

LightEMMA: A Longitudinal Evaluation of Vision-Language Models for Autonomous Driving

Rapid advances in vision-language models (VLMs) have generated growing interest in their application to autonomous driving. A prevailing assumption is that successive VLM generations will continually improve driving performance and eventually outperform state-of-the-art methods. To systematically examine this assumption, we introduce LightEMMA, a longitudinal framework for evaluating the autonomous driving performance of VLMs. LightEMMA uses a lightweight, unified evaluation protocol that assesses each model's intrinsic driving capability without model-specific fine-tuning, architectural changes, or prompt engineering. Using this protocol, we evaluate 15 models from five major families on the challenging nuScenes prediction benchmark. Empirical findings show that, despite increased model scale and enhanced general reasoning capabilities, successive VLM generations do not consistently achieve better driving performance. Further analysis of driving scenarios reveals recurring failure modes, including overreliance on historical actions and difficulty reconciling conflicting visual cues. These findings highlight the need for domain-specific adaptation to improve the safety of VLM-based autonomous driving systems. The source code is available at https://github.com/michigan-traffic-lab/LightEMMA.

cs.RO↗

arXiv2025

Quantum Annealing Algorithms for Estimating Ising Partition Functions

Estimating partition functions of Ising spin glasses is a cornerstone of statistical physics and computational science, yet it remains classically challenging due to its $\#$P-hard complexity. While Jarzynski's equality offers a theoretical pathway, its practical application is crippled at low temperatures by rare, divergent statistical fluctuations. Here, we introduce a quantum protocol that overcomes this fundamental limitation by synergizing reverse quantum annealing with optimized nonequilibrium initial distributions. Our method dramatically suppresses the estimator variance, achieving saturation in the low-temperature regime where existing methods fail. Numerical benchmarks on the Sherrington-Kirkpatrick spin glass and the 3-SAT problem demonstrate that our protocol reduces computational scaling exponents by over an order of magnitude (e.g., from $\sim 8.5$ to $\sim 0.5$), despite retaining exponential system-size dependence. Crucially, our protocol circumvents stringent adiabatic constraints, making it feasible for near-term quantum devices like superconducting qubits, trapped ions, and Rydberg atom arrays. This work provides a methodological framework for quantum-enhanced estimation in spin glass thermodynamics and beyond by harnessing non-adiabatic quantum dynamics to address a classically difficult problem.

quant-ph↗

arXiv2025

Towards Ambiguity-Free Spatial Foundation Model: Rethinking and Decoupling Depth Ambiguity

Depth ambiguity is a fundamental challenge in spatial scene understanding, especially in transparent scenes where single-depth estimates fail to capture full 3D structure. Existing models, limited to deterministic predictions, overlook real-world multi-layer depth. To address this, we introduce a paradigm shift from single-prediction to multi-hypothesis spatial foundation models. We first present \texttt{MD-3k}, a benchmark exposing depth biases in expert and foundational models through multi-layer spatial relationship labels and new metrics. To resolve depth ambiguity, we propose Laplacian Visual Prompting (LVP), a training-free spectral prompting technique that extracts hidden depth from pre-trained models via Laplacian-transformed RGB inputs. By integrating LVP-inferred depth with standard RGB-based estimates, our approach elicits multi-layer depth without model retraining. Extensive experiments validate the effectiveness of LVP in zero-shot multi-layer depth estimation, unlocking more robust and comprehensive geometry-conditioned visual generation, 3D-grounded spatial reasoning, and temporally consistent video-level depth inference. Our benchmark and code will be available at https://github.com/Xiaohao-Xu/Ambiguity-in-Space.

cs.CV↗

arXiv2025

Symmetry-induced fragmentation and dissipative time crystal

Time crystals are a peculiar state of matter. Their emergence hinges on ergodicity breaking, which typically originates from many-body localization or Floquet prethermalization. Here we propose a novel scheme for devising robust dissipative time crystals where the ergodicity is broken through symmetry-induced fragmentation. Building upon a U(1)-symmetry-induced Liouville-space fragmentation, we first propose a generic Liouvillian with long-time oscillations typical of time crystals. We then show that, even when the U(1) symmetry is broken, a prethermal time-crystal behavior survives, with distinct oscillation frequencies at different times of the steady-state approaching dynamics. Intriguingly, the stage-wise prethermal dynamics derive from Fermi statistics and the Liouvillian skin effect of our model -- as the excitations above the boundary-localized dark states can be mapped to the irreducible representations of the permutation group, the branching rules of the permutation group ensure the robustness of the prethermal time crystal. Our work paves the way for devising time crystals through Hilbert-space fragmentation. It also sheds light on the dynamic effects of non-Hermitian physics in many-body quantum open systems.

quant-ph↗

arXiv2024

FedFQ: Federated Learning with Fine-Grained Quantization

Federated learning (FL) is a decentralized approach, enabling multiple participants to collaboratively train a model while ensuring the protection of data privacy. The transmission of updates from numerous edge clusters to the server creates a significant communication bottleneck in FL. Quantization is an effective compression technology, showcasing immense potential in addressing this bottleneck problem. The Non-IID nature of FL renders it sensitive to quantization. Existing quantized FL frameworks inadequately balance high compression ratios and superior convergence performance by roughly employing a uniform quantization bit-width on the client-side. In this work, we propose a communication-efficient FL algorithm with a fine-grained adaptive quantization strategy (FedFQ). FedFQ addresses the trade-off between achieving high communication compression ratios and maintaining superior convergence performance by introducing parameter-level quantization. Specifically, we have designed a Constraint-Guided Simulated Annealing algorithm to determine specific quantization schemes. We derive the convergence of FedFQ, demonstrating its superior convergence performance compared to existing quantized FL algorithms. We conducted extensive experiments on multiple benchmarks and demonstrated that, while maintaining lossless performance, FedFQ achieves a compression ratio of 27 times to 63 times compared to the baseline experiment.

cs.DC↗

arXiv2024

Reducing Spurious Correlation for Federated Domain Generalization

The rapid development of multimedia has provided a large amount of data with different distributions for visual tasks, forming different domains. Federated Learning (FL) can efficiently use this diverse data distributed on different client media in a decentralized manner through model sharing. However, in open-world scenarios, there is a challenge: global models may struggle to predict well on entirely new domain data captured by certain media, which were not encountered during training. Existing methods still rely on strong statistical correlations between samples and labels to address this issue, which can be misleading, as some features may establish spurious short-cut correlations with the predictions. To comprehensively address this challenge, we introduce FedCD (Cross-Domain Invariant Federated Learning), an overall optimization framework at both the local and global levels. We introduce the Spurious Correlation Intervener (SCI), which employs invariance theory to locally generate interventers for features in a self-supervised manner to reduce the model's susceptibility to spurious correlated features. Our approach requires no sharing of data or features, only the gradients related to the model. Additionally, we develop the simple yet effective Risk Extrapolation Aggregation strategy (REA), determining aggregation coefficients through mathematical optimization to facilitate global causal invariant predictions. Extensive experiments and ablation studies highlight the effectiveness of our approach. In both classification and object detection generalization tasks, our method outperforms the baselines by an average of at least 1.45% in Acc, 4.8% and 1.27% in mAP50.

cs.CV↗

arXiv2024

Nearest is Not Dearest: Towards Practical Defense against Quantization-conditioned Backdoor Attacks

Model quantization is widely used to compress and accelerate deep neural networks. However, recent studies have revealed the feasibility of weaponizing model quantization via implanting quantization-conditioned backdoors (QCBs). These special backdoors stay dormant on released full-precision models but will come into effect after standard quantization. Due to the peculiarity of QCBs, existing defenses have minor effects on reducing their threats or are even infeasible. In this paper, we conduct the first in-depth analysis of QCBs. We reveal that the activation of existing QCBs primarily stems from the nearest rounding operation and is closely related to the norms of neuron-wise truncation errors (i.e., the difference between the continuous full-precision weights and its quantized version). Motivated by these insights, we propose Error-guided Flipped Rounding with Activation Preservation (EFRAP), an effective and practical defense against QCBs. Specifically, EFRAP learns a non-nearest rounding strategy with neuron-wise error norm and layer-wise activation preservation guidance, flipping the rounding strategies of neurons crucial for backdoor effects but with minimal impact on clean accuracy. Extensive evaluations on benchmark datasets demonstrate that our EFRAP can defeat state-of-the-art QCB attacks under various settings. Code is available at https://github.com/AntigoneRandy/QuantBackdoor_EFRAP.

cs.CR↗