Searcharxiv⌕ Search

arXiv subjects

Yuanzhe Li

Publications and source records attributed to Yuanzhe Li.

At least 19 recordsLinked to original sources

Quantization-Robust Unlearning through the Lens of Retain-Forget Loss Landscapes Interaction

Unlearning ensures LLM compliance by removing the influence of private or copyrighted training data. However, since LLM models typically undergo post-training compression, like quantization, in practical deployment, it has been observed that the unlearning effect can be substantially weakened, with the forgetting behavior degrading more severely than that of model utility. This paper proposes a quantization-robust unlearning framework that makes forgetting robust to quantization while maintaining overall model utility. We analyze this gap through the lens of loss landscape. Specifically, our analysis reveals a curvature-based criteria that pinpoints sensitive weights in the unlearned model that leads to both non-robust forgetting and reduced utility. We therefore propose sensitivity-guided noisy regularization, which is applied on the sensitive parameters to steer the model convergence towards a smoother minima of uniformly low forget and retain losses. Balancing unlearning and utility, we further propose forget-critical optimization, which updates only forget-critical layers, preserving most of the network to retain useful knowledge. Extensive experiments on the MUSE and TOFU benchmarks across multiple LLM unlearning algorithms show that our approach achieves substantially more quantization-resilient forgetting while maintaining utility.

cs.LG↗

Lightweight Pedestrian Head-Orientation Recognition Network for Safe Pedestrian-Vehicle Interaction

Pedestrian head orientation recognition plays an important role in autonomous driving by providing valuable cues for understanding pedestrian attention and anticipating potential crossing behavior. However, reliable recognition in real-world traffic scenes remains challenging because pedestrian head regions are often captured at low resolution. To address this challenge, we propose a lightweight Low-Resolution Head Orientation Convolutional Neural Network (LRHO-CNN) for pedestrian head orientation recognition. We construct a new dataset by extracting pedestrian head images from multiple public datasets and manually annotating them into eight orientation categories. The collected images are systematically preprocessed and augmented to increase data diversity and better represent variations in illumination and image quality. The experimental analysis compares LRHO-CNN with three fine-tuned baseline models, namely ResNet-18, ResNet-34, and VGG-16. The results demonstrate that LRHO-CNN achieves the highest classification accuracy among the evaluated models. LRHO-CNN is further evaluated on the JAAD and PIE datasets, demonstrating its effectiveness in recognizing pedestrian head orientation in real-world traffic scenes and providing informative head-orientation cues that can support downstream pedestrian behavior and intention prediction.

cs.CV↗

Detecting Phone-Induced Pedestrian Distraction via a Multimodal Fusion Transformer

The increasing reliance on mobile phones has made phone-induced pedestrian distraction increasingly prevalent. Activities such as texting, watching videos, and making phone calls have become significant contributors to traffic accidents. Reliable detection of pedestrian distraction is essential for autonomous vehicles, as it improves situational awareness and enables timely risk assessment, thereby supporting safe motion planning and vehicle control. We propose a multimodal fusion Transformer (MFT) for detecting phone-induced pedestrian distraction. MFT jointly extracts skeletal dynamics from body pose keypoints and visual appearance features from pedestrian images, effectively leveraging the complementary information provided by the two modalities. A cross-modal attention module is proposed to capture inter-modal dependencies through multi-head cross-attention, facilitating effective fusion of complementary information across the two modalities. Then, a temporal attention fusion module, implemented with a Transformer encoder, is employed to capture temporal dependencies. MFT is trained and evaluated on a manually annotated dataset comprising 287 pedestrian instances with 20,741 images. Extensive experiments demonstrate that MFT attains an overall accuracy of 95%, exceeding the performance of six baseline approaches by 6%.

cs.CV↗

Multi-Context Fusion Transformer for Pedestrian Crossing Intention Prediction in Urban Environments

Pedestrian crossing intention prediction is essential for autonomous vehicles to improve pedestrian safety and reduce traffic accidents. However, accurate pedestrian intention prediction in urban environments remains challenging due to the multitude of factors affecting pedestrian behavior. In this paper, we propose a multi-context fusion Transformer (MFT) that leverages diverse numerical contextual attributes across four key dimensions, encompassing pedestrian behavior context, environmental context, pedestrian localization context and vehicle motion context, to enable accurate pedestrian intention prediction. MFT employs a progressive fusion strategy, where mutual intra-context attention enables reciprocal interactions within each context, thereby facilitating feature sequence fusion and yielding a context token as a context-specific representation. This is followed by mutual cross-context attention, which integrates features across contexts with a global CLS token serving as a compact multi-context representation. Finally, guided intra-context attention refines context tokens within each context through directed interactions, while guided cross-context attention strengthens the global CLS token to promote multi-context fusion via guided information propagation, yielding deeper and more efficient integration. Experimental results validate the superiority of MFT over state-of-the-art methods, achieving accuracy rates of 73%, 93%, and 90% on the JAADbeh, JAADall, and PIE datasets, respectively. Extensive ablation studies are further conducted to investigate the effectiveness of the network architecture and contribution of different input context. Our code is open-source: https://github.com/ZhongHang0307/Multi-Context-Fusion-Transformer.

cs.CV↗

Outstanding TC Enhancement in 5d-3d Y2NiIrO6 by Compression

Understanding and predicting the properties of 5d compounds critically depend on the identification of the superexchange interactions from which their magnetism emerges. The study of pressure effects on double perovskite Y2NiIrO6 (YNIO) provides deep insight toward this goal. At ambient pressure, YNIO is a ferrimagnetic insulator with the Ir4+-5d Jeff = 1/2 Mott-insulating state. Under physical pressure up to 17 GPa, the compound exhibits concurrent compression on Ni/Ir-O bond lengths and Ni-O-Ir bond angles, leading to an increase of the Curie temperature from 192 to 243 K. On the contrary, external pressure increases distanced Ir-Ir interaction and in turn induces magnetic frustration in Sr2IrO4/Sr3Ir2O7 due to the extended 5d orbitals. In YNIO, the rock-salt ordered Ni-Ir naturally blocks extended superexchange beyond the nearest neighbor, and in turn suppresses such magnetic frustration. Moreover, the orthogonal Ni eg-Ir t2g pathway in YNIO is robust under lattice distortion, while the superexchange is weakened by bond bending in La2NiMnO6 with a similar half-filed eg-t2g configuration. Our findings establish a framework for elucidating the mechanism of 5d-3d superexchange and guide bond-engineered magnetism in iridate-related systems.

cond-mat.str-el↗

SAF3R: Dynamic Sparse Attention for Feed-Forward 3D Reconstruction Transformers

Feed-forward 3D reconstruction (F3R) transformers have recently achieved remarkable success. However, scaling them to long image sequences remains challenging, as the quadratic complexity of cross-view global attention quickly becomes the dominant computational bottleneck. While recent efforts attempt to improve efficiency through compressed or sparse attention, they fail to fully exploit the inherent sparsity and dynamic behavior of global attention. In this work, we present a comprehensive analysis of global attention across multiple F3R transformers and reveal that attention patterns are highly heterogeneous, dynamic, and extremely sparse across layers and attention heads. Motivated by these findings, we propose SAF3R, a training-free dynamic sparse attention framework tailored to F3R transformers. SAF3R integrates tailored sparse attention mechanisms with offline head profiling and an efficient online adaptation strategy to match input-dependent attention behaviors. Extensive experiments demonstrate that SAF3R achieves high sparsity ratios while preserving camera pose estimation and 3D reconstruction quality, translating into substantial end-to-end speedup on F3R transformers compared to existing methods. Code is available at https://github.com/jndeng/SAF3R

cs.CV↗

Dynamic Mixed-Precision Routing for Efficient Multi-step LLM Interaction

Large language models (LLMs) achieve strong performance in long-horizon decision-making tasks through multi-step interaction and reasoning at test time. While practitioners commonly believe a higher task success rate necessitates the use of a larger and stronger LLM model, multi-step interaction with a large LLM incurs prohibitive inference cost. To address this problem, we explore the use of low-precision quantized LLMs in the long-horizon decision-making process. Based on the observation of diverse sensitivities among interaction steps, we propose Dynamic Mixed-Precision Routing (DMR), a framework that adaptively selects between high-precision and low-precision LLMs at each decision step. The router is trained via a two-stage pipeline, consisting of KL-divergence-based supervised learning that identifies precision-sensitive steps, followed by Group-Relative Policy Optimization (GRPO) to further improve task success rates. Experiments on ALFWorld and WebShop demonstrate that our approach achieves a strong accuracy-cost trade-off over single-precision baselines.

cs.AI↗

Pressure-Induced Superconducting-like Transition in the $\it d$-wave Altermagnet Candidate CsV$_2$Se$_2$O

Altermagnetism generates exchange-type spin splitting without net magnetization and, in its $\it d$-wave form, resembles the angular symmetry of unconventional $\it d$-wave superconductivity. Whether this correspondence bears directly on superconducting instabilities in real correlated materials remains open. Here we study the quasi-two-dimensional vanadium oxychalcogenide CsV$_2$Se$_2$O (CVSO), a square-net $\it d$-wave altermagnet candidate, through combined experimental and theoretical investigation of its lattice structure, electronic structure and transport properties. At ambient pressure, CVSO is a weakly insulating parent state with a density-wave-like anomaly near 100 K, and its bulk properties are most consistent with a G-type compensated antiferromagnetic background. Under compression, the density-wave-like feature is suppressed, the magnetoresistance evolves from predominantly negative to positive, and a superconducting-like resistive downturn emerges below about 3 K. This low-temperature anomaly is reproducible across samples and pressure media, and is suppressed by magnetic field. Room-temperature X-ray diffraction reveals no symmetry lowering, whereas does show a pronounced compressibility anomaly over the same pressure range. CVSO thus reveals a pressure-tuned phase diagram in which a reconstructed weakly insulating parent state gives way to strange-metal-like transport and superconducting-like behavior, echoing broader phenomenology associated with unconventional superconductors, including cuprates and nickelates.

cond-mat.supr-con↗

Gastric-X: A Multimodal Multi-Phase Benchmark Dataset for Advancing Vision-Language Models in Gastric Cancer Analysis

Recent vision-language models (VLMs) have shown strong generalization and multimodal reasoning abilities in natural domains. However, their application to medical diagnosis remains limited by the lack of comprehensive and structured datasets that capture real clinical workflows. To advance the development of VLMs for clinical applications, particularly in gastric cancer, we introduce Gastric-X, a large-scale multimodal benchmark for gastric cancer analysis providing 1.7K cases. Each case in Gastric-X includes paired resting and dynamic CT scans, endoscopic image, a set of structured biochemical indicators, expert-authored diagnostic notes, and bounding box annotations of tumor regions, reflecting realistic clinical conditions. We systematically examine the capability of recent VLMs on five core tasks: Visual Question Answering (VQA), report generation, cross-modal retrieval, disease classification, and lesion localization. These tasks simulate critical stages of clinical workflow, from visual understanding and reasoning to multimodal decision support. Through this evaluation, we aim not only to assess model performance but also to probe the nature of VLM understanding: Can current VLMs meaningfully correlate biochemical signals with spatial tumor features and textual reports? We envision Gastric-X as a step toward aligning machine intelligence with the cognitive and evidential reasoning processes of physicians, and as a resource to inspire the development of next-generation medical VLMs.

cs.CV↗

Pedestrian Crossing Intention Prediction Using Multimodal Fusion Network

Pedestrian crossing intention prediction is essential for the deployment of autonomous vehicles (AVs) in urban environments. Ideal prediction provides AVs with critical environmental cues, thereby reducing the risk of pedestrian-related collisions. However, the prediction task is challenging due to the diverse nature of pedestrian behavior and its dependence on multiple contextual factors. This paper proposes a multimodal fusion network that leverages seven modality features from both visual and motion branches, aiming to effectively extract and integrate complementary cues across different modalities. Specifically, motion and visual features are extracted from the raw inputs using multiple Transformer-based extraction modules. Depth-guided attention module leverages depth information to guide attention towards salient regions in another modality through comprehensive spatial feature interactions. To account for the varying importance of different modalities and frames, modality attention and temporal attention are designed to selectively emphasize informative modalities and effectively capture temporal dependencies. Extensive experiments on the JAAD dataset validate the effectiveness of the proposed network, achieving superior performance compared to the baseline methods.

cs.CV↗

Doracamom: Joint 3D Detection and Occupancy Prediction with Multi-view 4D Radars and Cameras for Omnidirectional Perception

3D object detection and occupancy prediction are critical tasks in autonomous driving, attracting significant attention. Despite the potential of recent vision-based methods, they encounter challenges under adverse conditions. Thus, integrating cameras with next-generation 4D imaging radar to achieve unified multi-task perception is highly significant, though research in this domain remains limited. In this paper, we propose Doracamom, the first framework that fuses multi-view cameras and 4D radar for joint 3D object detection and semantic occupancy prediction, enabling comprehensive environmental perception. Specifically, we introduce a novel Coarse Voxel Queries Generator that integrates geometric priors from 4D radar with semantic features from images to initialize voxel queries, establishing a robust foundation for subsequent Transformer-based refinement. To leverage temporal information, we design a Dual-Branch Temporal Encoder that processes multi-modal temporal features in parallel across BEV and voxel spaces, enabling comprehensive spatio-temporal representation learning. Furthermore, we propose a Cross-Modal BEV-Voxel Fusion module that adaptively fuses complementary features through attention mechanisms while employing auxiliary tasks to enhance feature quality. Extensive experiments on the OmniHD-Scenes, View-of-Delft (VoD), and TJ4DRadSet datasets demonstrate that Doracamom achieves state-of-the-art performance in both tasks, establishing new benchmarks for multi-modal 3D perception. Code and models will be publicly available.

cs.CV↗

AIConfigurator: Lightning-Fast Configuration Optimization for Multi-Framework LLM Serving

Optimizing Large Language Model (LLM) inference in production systems is increasingly difficult due to dynamic workloads, stringent latency/throughput targets, and a rapidly expanding configuration space. This complexity spans not only distributed parallelism strategies (tensor/pipeline/expert) but also intricate framework-specific runtime parameters such as those concerning the enablement of CUDA graphs, available KV-cache memory fractions, and maximum token capacity, which drastically impact performance. The diversity of modern inference frameworks (e.g., TRT-LLM, vLLM, SGLang), each employing distinct kernels and execution policies, makes manual tuning both framework-specific and computationally prohibitive. We present AIConfigurator, a unified performance-modeling system that enables rapid, framework-agnostic inference configuration search without requiring GPU-based profiling. AIConfigurator combines (1) a methodology that decomposes inference into analytically modelable primitives - GEMM, attention, communication, and memory operations while capturing framework-specific scheduling dynamics; (2) a calibrated kernel-level performance database for these primitives across a wide range of hardware platforms and popular open-weights models (GPT-OSS, Qwen, DeepSeek, LLama, Mistral); and (3) an abstraction layer that automatically resolves optimal launch parameters for the target backend, seamlessly integrating into production-grade orchestration systems. Evaluation on production LLM serving workloads demonstrates that AIConfigurator identifies superior serving configurations that improve performance by up to 40% for dense models (e.g., Qwen3-32B) and 50% for MoE architectures (e.g., DeepSeek-V3), while completing searches within 30 seconds on average. Enabling the rapid exploration of vast design spaces - from cluster topology down to engine specific flags.

cs.LG↗

Programmable calculus operations in electromagnetic space using space-time-coding metasurface

With the rapid advancement of metasurfaces and the increasing demand for programmable metasurfaces to simplify information systems, wave-based computation using metasurfaces has emerged as an attractive research topic. To facilitate the mathematical operations in electromagnetic (EM) space, here we propose a space-time coding metasurface (STCM) system capable of directly performing calculus operations on the spatial energy distributions of EM waves. By exploiting harmonic characteristics induced by time-varying coding, the responses of meta-atoms at specific harmonics can be flexibly controlled, which enables the metasurface system to address more complex tasks. Owing to its programmability, the STCM can dynamically switch functions in real time to accommodate different calculus tasks. To fully leverage the capability of STCM, we not only present the space-time coding sequences for differentiation and integration of EM waves, but also develop and numerically simulate the space-time coding sequences that can independently and simultaneously implement different calculus operations on the same incident EM waves. To experimentally validate the feasibility of the EM calculus operations, proof-of-concept experiments are conducted using a programmable 2-bit STCM. Good agreements among the theory, numerical simulations, and experiments confirm the feasibility of performing calculus operations in the EM space and demonstrate the broad application prospects of STCM in EM wave manipulations, wireless communications, and signal processing.

physics.optics↗

Disorder-mediated linear and nonlinear magnetotransport in the charge-density-wave material ${\rm Ta_2NiSe_7 }$

We report disorder-mediated first-order linear and higher-order nonlinear (magneto-)transport of Ta$_2$NiSe$_7$ (TNS) in the charge-density-wave (CDW) regime. CDW transition temperature ($T_{CDW}$) and carrier density are proportional and inversely proportional to residual resistance ratio of samples, respectively. Such relation helps to understand the unique CDW order therein. High-$T_{CDW}$ TNS exhibits negative first-harmonic magnetoresistance (MR$^{1ω}$) under a magnetic field ($B$) parallel to the direction of alternating current ($I^ω$), which may arise from the anomalous velocity induced by the Berry curvature of three-dimensional topological bands near the Fermi level. As $T_{CDW}$ drops, a positive-to-negative MR$^{1ω}$ transition is observed with decreasing perpendicular $B$, which is likely due to the contribution of Zeeman effect on current pathways in the disordered system. Moreover, interestingly, the second-harmonic nonlinear signals are suppressed, while the third-harmonic signals are significant and sensitive to both $B$ and $T_{CDW}$. Such observations, together with scaling analysis, suggest the quantum geometry quadrupole at play and the modulation of disorder on third-order nonlinearity. Our results pave an avenue for tailoring distinct-order magnetoresistive phases in disordered topological materials.

cond-mat.mtrl-sci↗

ACIT: Attention-Guided Cross-Modal Interaction Transformer for Pedestrian Crossing Intention Prediction

Predicting pedestrian crossing intention is crucial for autonomous vehicles to prevent pedestrian-related collisions. However, effectively extracting and integrating complementary cues from different types of data remains one of the major challenges. This paper proposes an attention-guided cross-modal interaction Transformer (ACIT) for pedestrian crossing intention prediction. ACIT leverages six visual and motion modalities, which are grouped into three interaction pairs: (1) Global semantic map and global optical flow, (2) Local RGB image and local optical flow, and (3) Ego-vehicle speed and pedestrian's bounding box. Within each visual interaction pair, a dual-path attention mechanism enhances salient regions within the primary modality through intra-modal self-attention and facilitates deep interactions with the auxiliary modality (i.e., optical flow) via optical flow-guided attention. Within the motion interaction pair, cross-modal attention is employed to model the cross-modal dynamics, enabling the effective extraction of complementary motion features. Beyond pairwise interactions, a multi-modal feature fusion module further facilitates cross-modal interactions at each time step. Furthermore, a Transformer-based temporal feature aggregation module is introduced to capture sequential dependencies. Experimental results demonstrate that ACIT outperforms state-of-the-art methods, achieving accuracy rates of 70% and 89% on the JAADbeh and JAADall datasets, respectively. Extensive ablation studies are further conducted to investigate the contribution of different modules of ACIT.

cs.CV↗

SimLOB: Learning Representations of Limited Order Book for Financial Market Simulation

Financial market simulation (FMS) serves as a promising tool for understanding market anomalies and the underlying trading behaviors. To ensure high-fidelity simulations, it is crucial to calibrate the FMS model for generating data closely resembling the observed market data. Previous efforts primarily focused on calibrating the mid-price data, leading to essential information loss of the market activities and thus biasing the calibrated model. The Limit Order Book (LOB) data is the fundamental data fully capturing the market micro-structure and is adopted by worldwide exchanges. However, LOB is not applicable to existing calibration objective functions due to its tabular structure not suitable for the vectorized input requirement. This paper proposes to explicitly learn the vectorized representations of LOB with a Transformer-based autoencoder. Then the latent vector, which captures the major information of LOB, can be applied for calibration. Extensive experiments show that the learned latent representation not only preserves the non-linear auto-correlation in the temporal axis, but the precedence between successive price levels of LOB. Besides, it is verified that the performance of the representation learning stage is consistent with the downstream calibration tasks. Thus, this work also progresses the FMS on LOB data, for the first time.

cs.CE↗

Inverse Design of Metamaterials with Manufacturing-Guiding Spectrum-to-Structure Conditional Diffusion Model

Metamaterials are artificially engineered structures that manipulate electromagnetic waves, having optical properties absent in natural materials. Recently, machine learning for the inverse design of metamaterials has drawn attention. However, the highly nonlinear relationship between the metamaterial structures and optical behaviour, coupled with fabrication difficulties, poses challenges for using machine learning to design and manufacture complex metamaterials. Herein, we propose a general framework that implements customised spectrum-to-shape and size parameters to address one-to-many metamaterial inverse design problems using conditional diffusion models. Our method exhibits superior spectral prediction accuracy, generates a diverse range of patterns compared to other typical generative models, and offers valuable prior knowledge for manufacturing through the subsequent analysis of the diverse generated results, thereby facilitating the experimental fabrication of metamaterial designs. We demonstrate the efficacy of the proposed method by successfully designing and fabricating a free-form metamaterial with a tailored selective emission spectrum for thermal camouflage applications.

physics.optics↗

Bridge-Coder: Unlocking LLMs' Potential to Overcome Language Gaps in Low-Resource Code

Large Language Models (LLMs) demonstrate strong proficiency in generating code for high-resource programming languages (HRPLs) like Python but struggle significantly with low-resource programming languages (LRPLs) such as Racket or D. This performance gap deepens the digital divide, preventing developers using LRPLs from benefiting equally from LLM advancements and reinforcing disparities in innovation within underrepresented programming communities. While generating additional training data for LRPLs is promising, it faces two key challenges: manual annotation is labor-intensive and costly, and LLM-generated LRPL code is often of subpar quality. The underlying cause of this issue is the gap between natural language to programming language gap (NL-PL Gap), which is especially pronounced in LRPLs due to limited aligned data. In this work, we introduce a novel approach called Bridge-Coder, which leverages LLMs' intrinsic capabilities to enhance the performance on LRPLs. Our method consists of two key stages. Bridge Generation, where we create high-quality dataset by utilizing LLMs' general knowledge understanding, proficiency in HRPLs, and in-context learning abilities. Then, we apply the Bridged Alignment, which progressively improves the alignment between NL instructions and LRPLs. Experimental results across multiple LRPLs show that Bridge-Coder significantly enhances model performance, demonstrating the effectiveness and generalization of our approach. Furthermore, we offer a detailed analysis of the key components of our method, providing valuable insights for future work aimed at addressing the challenges associated with LRPLs.

cs.CL↗