arXiv subjects

Hongyi Li

Publications and source records attributed to Hongyi Li.

At least 37 records · Page 2Linked to original sources

arXiv2026

Rotation-Robust Regression with Convolutional Model Trees

We study rotation-robust learning for image inputs using Convolutional Model Trees (CMTs) [1], whose split and leaf coefficients can be structured on the image grid and transformed geometrically at deployment time. In a controlled MNIST setting with a rotation-invariant regression target, we introduce three geometry-aware inductive biases for split directions -- convolutional smoothing, a tilt dominance constraint, and importance-based pruning -- and quantify their impact on robustness under in-plane rotations. We further evaluate a deployment-time orientation search that selects a discrete rotation maximizing a forest-level confidence proxy without updating model parameters. Orientation search improves robustness under severe rotations but can be harmful near the canonical orientation when confidence is misaligned with correctness. Finally, we observe consistent trends on MNIST digit recognition implemented as one-vs-rest regression, highlighting both the promise and limitations of confidence-based orientation selection for model-tree ensembles.

cs.CV↗

arXiv2025

START: Traversing Sparse Footholds with Terrain Reconstruction

Traversing terrains with sparse footholds like legged animals presents a promising yet challenging task for quadruped robots, as it requires precise environmental perception and agile control to secure safe foot placement while maintaining dynamic stability. Model-based hierarchical controllers excel in laboratory settings, but suffer from limited generalization and overly conservative behaviors. End-to-end learning-based approaches unlock greater flexibility and adaptability, but existing state-of-the-art methods either rely on heightmaps that introduce noise and complex, costly pipelines, or implicitly infer terrain features from egocentric depth images, often missing accurate critical geometric cues and leading to inefficient learning and rigid gaits. To overcome these limitations, we propose START, a single-stage learning framework that enables agile, stable locomotion on highly sparse and randomized footholds. START leverages only low-cost onboard vision and proprioception to accurately reconstruct local terrain heightmap, providing an explicit intermediate representation to convey essential features relevant to sparse foothold regions. This supports comprehensive environmental understanding and precise terrain assessment, reducing exploration cost and accelerating skill acquisition. Experimental results demonstrate that START achieves zero-shot transfer across diverse real-world scenarios, showcasing superior adaptability, precise foothold placement, and robust locomotion.

cs.RO↗

arXiv2025

StyleBreak: Revealing Alignment Vulnerabilities in Large Audio-Language Models via Style-Aware Audio Jailbreak

Large Audio-language Models (LAMs) have recently enabled powerful speech-based interactions by coupling audio encoders with Large Language Models (LLMs). However, the security of LAMs under adversarial attacks remains underexplored, especially through audio jailbreaks that craft malicious audio prompts to bypass alignment. Existing efforts primarily rely on converting text-based attacks into speech or applying shallow signal-level perturbations, overlooking the impact of human speech's expressive variations on LAM alignment robustness. To address this gap, we propose StyleBreak, a novel style-aware audio jailbreak framework that systematically investigates how diverse human speech attributes affect LAM alignment robustness. Specifically, StyleBreak employs a two-stage style-aware transformation pipeline that perturbs both textual content and audio to control linguistic, paralinguistic, and extralinguistic attributes. Furthermore, we develop a query-adaptive policy network that automatically searches for adversarial styles to enhance the efficiency of LAM jailbreak exploration. Extensive evaluations demonstrate that LAMs exhibit critical vulnerabilities when exposed to diverse human speech attributes. Moreover, StyleBreak achieves substantial improvements in attack effectiveness and efficiency across multiple attack paradigms, highlighting the urgent need for more robust alignment in LAMs.

cs.SD↗

arXiv2025

Underground Power Distribution System Restoration Using Inverter Based Resources

Underground power distribution systems (PDSs) are increasingly deployed in urban areas. The integration of smart devices including smart switchgears, pad-mounted distribution transformers and inverter-based resources (IBRs) enhance system resilience, however simultaneously introducing unique challenges. The challenges include inrush currents caused by trapped charges in underground cables, ferroresonance in distribution transformers during energization, and three-phase load imbalance resulting from single-phase underground laterals. To address these issues, this paper proposes an underground PDS restoration framework using IBRs. Firstly, an underground cable energization model is developed to quantify inrush current by analyzing voltage differences across both switchgear terminals. Secondly, a distribution transformer energization model is proposed to evaluate ferroresonance using Q-factor constraints based on underground cable capacitance and damping resistance. Thirdly, a phase-swapping model is proposed to improve load balancing by dynamically reassigning lateral-phase connections through smart switchgears. The proposed models are further integrated into a mixed-integer nonlinear programming (MINLP) formulation to maximize the total weighted restored load while constraining inrush currents, ferroresonance, and phase imbalance. To address the nonlinearity induced by impedance matrix reordering during phase swapping, a permutation-based linearization technique is proposed. Finally, case studies on an underground PDS established based on IEEE 123-Node Test Feeder validate the effectiveness of the proposed strategy in improving uderground PDS restoration performance.

eess.SY↗

arXiv2025

Data-Driven Stochastic Distribution System Hardening Based on Bayesian Online Learning

Extreme weather frequently cause widespread outages in distribution systems (DSs), demonstrating the importance of hardening strategies for resilience enhancement. However, the well-utilization of real-world outage data with associated weather conditions to make informed hardening decisions in DSs is still an open issue. To bridge this research gap, this paper proposes a data-driven stochastic distribution line (DL) hardening strategy. First, a deep neural network (DNN) regression model is developed to predict the probabilistic evolution of outage scenarios under various hardening decisions. Based on the DNN predictions, the problem is formulated as a decision-dependent distributionally robust optimization (DRO) model, accounting for uncertainties in outage scenario distributions using a data-driven ambiguity set. To address decision-dependent uncertainty, a Bayesian online learning algorithm is proposed. This algorithm decomposes the original problem into inner and outer problems. Then, it iteratively refines hardening decisions by sequentially incorporating outage data and dynamically updating decision-specific ambiguity sets by using Bayes' theorem and Bayesian Inference. Also, the convergence of the algorithm is proven through dynamic regret analysis. Finally, case studies are implemented on a real-world DS in Redfield, Iowa, USA. A dataset spanning 24 years (2001-2024) is constructed based on the utility outage records. The simulation results validates the effectiveness of the proposed strategy.

eess.SY↗

arXiv2025

Power Distribution System Blackstart Restoration Using Renewable Energy

Integrating renewable energy sources into the grid not only reduces global carbon emissions, but also facilitates distribution system (DS) blackstart restoration. This process leverages renewable energy, inverters, situational awareness and distribution automation to initiate blackstart at the DS level, obtaining a fast response and bottom-up restoration. In this Review, we survey the latest technological advances for DS blackstart restoration using renewable energy. We first present mathematical models for distributed energy resources (DERs), network topology, and load dynamics. We then discuss how the situational awareness can help improve restoration performance through real-time monitoring and forecasting. Next, the DS blackstart restoration problem, including objectives, constraints, and existing methodologies for decision-making are provided. Lastly, we outline remaining challenges, and highlight the opportunities and future research directions.

eess.SY↗

arXiv2025

Light-SQ: Structure-aware Shape Abstraction with Superquadrics for Generated Meshes

In user-generated-content (UGC) applications, non-expert users often rely on image-to-3D generative models to create 3D assets. In this context, primitive-based shape abstraction offers a promising solution for UGC scenarios by compressing high-resolution meshes into compact, editable representations. Towards this end, effective shape abstraction must therefore be structure-aware, characterized by low overlap between primitives, part-aware alignment, and primitive compactness. We present Light-SQ, a novel superquadric-based optimization framework that explicitly emphasizes structure-awareness from three aspects. (a) We introduce SDF carving to iteratively udpate the target signed distance field, discouraging overlap between primitives. (b) We propose a block-regrow-fill strategy guided by structure-aware volumetric decomposition, enabling structural partitioning to drive primitive placement. (c) We implement adaptive residual pruning based on SDF update history to surpress over-segmentation and ensure compact results. In addition, Light-SQ supports multiscale fitting, enabling localized refinement to preserve fine geometric details. To evaluate our method, we introduce 3DGen-Prim, a benchmark extending 3DGen-Bench with new metrics for both reconstruction quality and primitive-level editability. Extensive experiments demonstrate that Light-SQ enables efficient, high-fidelity, and editable shape abstraction with superquadrics for complex generated geometry, advancing the feasibility of 3D UGC creation.

cs.GR↗

arXiv2025

CompAir: Synergizing Complementary PIMs and In-Transit NoC Computation for Efficient LLM Acceleration

The rapid advancement of Large Language Models (LLMs) has revolutionized various aspects of human life, yet their immense computational and energy demands pose significant challenges for efficient inference. The memory wall, the growing processor-memory speed disparity, remains a critical bottleneck for LLM. Process-In-Memory (PIM) architectures overcome limitations by co-locating compute units with memory, leveraging 5-20$\times$ higher internal bandwidth and enabling greater energy efficiency than GPUs. However, existing PIMs struggle to balance flexibility, performance, and cost-efficiency for LLMs' dynamic memory-compute patterns and operator diversity. DRAM-PIM suffers from inter-bank communication overhead despite its vector parallelism. SRAM-PIM offers sub-10ns latency for matrix operation but is constrained by limited capacity. This work introduces CompAir, a novel PIM architecture that integrates DRAM-PIM and SRAM-PIM with hybrid bonding, enabling efficient linear computations while unlocking multi-granularity data pathways. We further develop CompAir-NoC, an advanced network-on-chip with an embedded arithmetic logic unit that performs non-linear operations during data movement, simultaneously reducing communication overhead and area cost. Finally, we develop a hierarchical Instruction Set Architecture that ensures both flexibility and programmability of the hybrid PIM. Experimental results demonstrate that CompAir achieves 1.83-7.98$\times$ prefill and 1.95-6.28$\times$ decode improvement over the current state-of-the-art fully PIM architecture. Compared to the hybrid A100 and HBM-PIM system, CompAir achieves 3.52$\times$ energy consumption reduction with comparable throughput. This work represents the first systematic exploration of hybrid DRAM-PIM and SRAM-PIM architectures with in-network computation capabilities, offering a high-efficiency solution for LLM.

cs.AR↗

arXiv2025

A Synthetic Texas Power System with Time-Series Weather-Dependent Spatiotemporal Profiles

We developed a synthetic Texas 123-bus backbone transmission system (TX-123BT) with spatio-temporally correlated grid profiles of solar power, wind power, dynamic line ratings and loads at one-hour resolution for five continuous years, which demonstrates unique advantages compared to conventional test cases that offer single static system profile snapshots. Three weather-dependent models are used to create the hourly wind power productions, solar power productions, and dynamic line ratings respectively. The actual historical weather information is also provided along with this dataset, which is suitable for machine learning models. Security-constrained unit commitment is conducted on TX-123BT daily grid profiles and numerical results are compared with the actual Texas system for validation. The created hourly DLR profiles can cut operating cost from USD 8.09 M to USD 7.95 M (-1.7 %), raises renewable dispatch by 1.3 %, and lowers average LMPs from USD 18.66 to USD 17.98 /MWh (-3.6 %). Two hydrogen options -- a 200 MW dual hub and a 500 MW hydrogen-energy transmission and conversion system -- reduce high-load Q3 daily costs by 13.9 % and 14.1 %, respectively. Sensitivity tests show that suppressing the high-resolution weather-driven profiles can push system cost up by as much as 15 %, demonstrating the economic weight of temporal detail.

eess.SY↗

arXiv2025

SATA: Safe and Adaptive Torque-Based Locomotion Policies Inspired by Animal Learning

Despite recent advances in learning-based controllers for legged robots, deployments in human-centric environments remain limited by safety concerns. Most of these approaches use position-based control, where policies output target joint angles that must be processed by a low-level controller (e.g., PD or impedance controllers) to compute joint torques. Although impressive results have been achieved in controlled real-world scenarios, these methods often struggle with compliance and adaptability when encountering environments or disturbances unseen during training, potentially resulting in extreme or unsafe behaviors. Inspired by how animals achieve smooth and adaptive movements by controlling muscle extension and contraction, torque-based policies offer a promising alternative by enabling precise and direct control of the actuators in torque space. In principle, this approach facilitates more effective interactions with the environment, resulting in safer and more adaptable behaviors. However, challenges such as a highly nonlinear state space and inefficient exploration during training have hindered their broader adoption. To address these limitations, we propose SATA, a bio-inspired framework that mimics key biomechanical principles and adaptive learning mechanisms observed in animal locomotion. Our approach effectively addresses the inherent challenges of learning torque-based policies by significantly improving early-stage exploration, leading to high-performance final policies. Remarkably, our method achieves zero-shot sim-to-real transfer. Our experimental results indicate that SATA demonstrates remarkable compliance and safety, even in challenging environments such as soft/slippery terrain or narrow passages, and under significant external disturbances, highlighting its potential for practical deployments in human-centric and safety-critical scenarios.

cs.RO↗

arXiv2025

LHT: Statistically-Driven Oblique Decision Trees for Interpretable Classification

We introduce the Learning Hyperplane Tree (LHT), a novel oblique decision tree model designed for expressive and interpretable classification. LHT fundamentally distinguishes itself through a non-iterative, statistically-driven approach to constructing splitting hyperplanes. Unlike methods that rely on iterative optimization or heuristics, LHT directly computes the hyperplane parameters, which are derived from feature weights based on the differences in feature expectations between classes within each node. This deterministic mechanism enables a direct and well-defined hyperplane construction process. Predictions leverage a unique piecewise linear membership function within leaf nodes, obtained via local least-squares fitting. We formally analyze the convergence of the LHT splitting process, ensuring that each split yields meaningful, non-empty partitions. Furthermore, we establish that the time complexity for building an LHT up to depth $d$ is $O(mnd)$, demonstrating the practical feasibility of constructing trees with powerful oblique splits using this methodology. The explicit feature weighting at each split provides inherent interpretability. Experimental results on benchmark datasets demonstrate LHT's competitive accuracy, positioning it as a practical, theoretically grounded, and interpretable alternative in the landscape of tree-based models. The implementation of the proposed method is available at https://github.com/Hongyi-Li-sz/LHT_model.

cs.LG↗

arXiv2025

MLDSE: Scaling Design Space Exploration Infrastructure for Multi-Level Hardware

To efficiently support large-scale NNs, multi-level hardware, leveraging advanced integration and interconnection technologies, has emerged as a promising solution to counter the slowdown of Moore's law. However, the vast design space of such hardware, coupled with the complexity of their spatial hierarchies and organizations, introduces significant challenges for design space exploration (DSE). Existing DSE tools, which rely on predefined hardware templates to explore parameters for specific architectures, fall short in exploring diverse organizations, spatial hierarchies, and architectural polymorphisms inherent in multi-level hardware. To address these limitations, we present Multi-Level Design Space Exploror (MLDSE), a novel infrastructure for domain-specific DSE of multi-level hardware. MLDSE introduces three key innovations from three basic perspectives of DSE: 1) Modeling: MLDSE introduces a hardware intermediate representation (IR) that can recursively model diverse multi-level hardware with composable elements at various granularities. 2) Mapping: MLDSE provides a comprehensive spatiotemporal mapping IR and mapping primitives, facilitating the mapping strategy exploration on multi-level hardware, especially synchronization and cross-level communication; 3) Simulation: MLDSE supports universal simulator generation based on task-level event-driven simulation mechanism. It features a hardware-consistent scheduling algorithm that can handle general task-level resource contention. Through experiments on LLM workloads, we demonstrate MLDSE's unique capability to perform three-tier DSE spanning architecture, hardware parameter, and mapping.

cs.AR↗

arXiv2025

Differentially Private In-context Learning via Sampling Few-shot Mixed with Zero-shot Outputs

In-context learning (ICL) has shown promising improvement in downstream task adaptation of LLMs by augmenting prompts with relevant input-output examples (demonstrations). However, the ICL demonstrations can contain privacy-sensitive information, which can be leaked and/or regurgitated by the LLM output. Differential Privacy (DP), a widely adopted privacy safeguard, has emerged to mitigate this privacy leakage, with recent work demonstrating strong privacy-utility tradeoffs in classification tasks for ICL. However, generation tasks for ICL are challenging due to the high-dimensional output space of open-ended generation. To this end, we propose $\texttt{dps-mozo}$, Differentially Private Sampling by Mixing One-shot with Zero-shot Outputs, a decoding framework that generates DP text by sampling from the product of multiple one-shot outputs mixed with a zero-shot output. This mixing effectively reduces the amount of information that can be leaked by each demonstration. By utilizing the inherent randomness in sampling from the mixed distributions, we can achieve DP without adding noise, thereby improving the privacy-utility tradeoff. Our experimental evaluations show $\texttt{dps-mozo}$ can achieve a strong privacy guarantee, $ε=2$, with minimal utility degradation compared to non-private few-shot learning, $\textbf{0.3}$% ROUGE-L F1 score decrease on the SAMSum dataset with Gemma 2 2B.

cs.LG↗

arXiv2025

Learning Hyperplane Tree: A Piecewise Linear and Fully Interpretable Decision-making Framework

This paper introduces a novel tree-based model, Learning Hyperplane Tree (LHT), which outperforms state-of-the-art (SOTA) tree models for classification tasks on several public datasets. The structure of LHT is simple and efficient: it partitions the data using several hyperplanes to progressively distinguish between target and non-target class samples. Although the separation is not perfect at each stage, LHT effectively improves the distinction through successive partitions. During testing, a sample is classified by evaluating the hyperplanes defined in the branching blocks and traversing down the tree until it reaches the corresponding leaf block. The class of the test sample is then determined using the piecewise linear membership function defined in the leaf blocks, which is derived through least-squares fitting and fuzzy logic. LHT is highly transparent and interpretable--at each branching block, the contribution of each feature to the classification can be clearly observed.

cs.LG↗

arXiv2024

On Monotone Persuasion

We study monotone persuasion in the linear case, where posterior distributions over states are summarized by their mean. We solve the two leading cases where optimal unrestricted signals can be nonmonotone. First, if the objective is s-shaped and the state is discrete, then optimal monotone signals are upper censorship, whereas optimal unrestricted signals may require randomization. Second, if the objective is m-shaped and the state is continuous, then optimal monotone signals are interval disclosure, whereas optimal unrestricted signals may require nonmonotone pooling. We illustrate our results with an application to media censorship.

econ.TH↗

arXiv2024

JailPO: A Novel Black-box Jailbreak Framework via Preference Optimization against Aligned LLMs

Large Language Models (LLMs) aligned with human feedback have recently garnered significant attention. However, it remains vulnerable to jailbreak attacks, where adversaries manipulate prompts to induce harmful outputs. Exploring jailbreak attacks enables us to investigate the vulnerabilities of LLMs and further guides us in enhancing their security. Unfortunately, existing techniques mainly rely on handcrafted templates or generated-based optimization, posing challenges in scalability, efficiency and universality. To address these issues, we present JailPO, a novel black-box jailbreak framework to examine LLM alignment. For scalability and universality, JailPO meticulously trains attack models to automatically generate covert jailbreak prompts. Furthermore, we introduce a preference optimization-based attack method to enhance the jailbreak effectiveness, thereby improving efficiency. To analyze model vulnerabilities, we provide three flexible jailbreak patterns. Extensive experiments demonstrate that JailPO not only automates the attack process while maintaining effectiveness but also exhibits superior performance in efficiency, universality, and robustness against defenses compared to baselines. Additionally, our analysis of the three JailPO patterns reveals that attacks based on complex templates exhibit higher attack strength, whereas covert question transformations elicit riskier responses and are more likely to bypass defense mechanisms.

cs.CR↗

arXiv2024

General-purpose Dataflow Model with Neuromorphic Primitives

Neuromorphic computing exhibits great potential to provide high-performance benefits in various applications beyond neural networks. However, a general-purpose program execution model that aligns with the features of neuromorphic computing is required to bridge the gap between program versatility and neuromorphic hardware efficiency. The dataflow model offers a potential solution, but it faces high graph complexity and incompatibility with neuromorphic hardware when dealing with control flow programs, which decreases the programmability and performance. Here, we present a dataflow model tailored for neuromorphic hardware, called neuromorphic dataflow, which provides a compact, concise, and neuromorphic-compatible program representation for control logic. The neuromorphic dataflow introduces "when" and "where" primitives, which restructure the view of control. The neuromorphic dataflow embeds these primitives in the dataflow schema with the plasticity inherited from the spiking algorithms. Our method enables the deployment of general-purpose programs on neuromorphic hardware with both programmability and plasticity, while fully utilizing the hardware's potential.

cs.CL↗

arXiv2024

Priority-based Energy Allocation in Buildings through Distributed Model Predictive Control

Many countries are facing energy shortage today and most of the global energy is consumed by HVAC systems in buildings. For the scenarios where the energy system is not sufficiently supplied to HVAC systems, a priority-based allocation scheme based on distributed model predictive control is proposed in this paper, which distributes the energy rationally based on priority order. According to the scenarios, two distributed allocation strategies, i.e., one-to-one priority strategy and multi-to-one priority strategy, are developed in this paper and validated by simulation in a building containing three zones and a building containing 36 rooms, respectively. Both priority-based strategies fully exploit the potential of predictive control solutions. The experiment shows that our scheme has good scalability and achieve the performance of centralized strategy while making the calculation tractable.

eess.SY↗