Searcharxiv⌕ Search

arXiv subjects

Yi Qian

Publications and source records attributed to Yi Qian.

At least 19 recordsLinked to original sources
arXiv2026

Open-Source Autonomous Driving System Analysis and Multi-Disciplinary Hardware-in-the-Loop Research Paradigm with Reinforcement-Learning Testing and Large Language Models

Open-source autonomous driving systems provide an inspectable software foundation for intelligent vehicle research. Under real-vehicle deployment conditions, the recording and review of experimental conditions are important for interpreting system behavior and reusing experimental results. However, in a shared real-vehicle environment involving multiple vehicles, task processes, code modifications, and hardware testing feedback are often distributed across different teams and experimental stages, making it challenging to maintain continuous and reviewable experimental records. To address this limitation, this paper examines an Apollo-on-Hongqi EV environment and proposes a real-vehicle experimental framework. The framework connects multi-vehicle experiments, repository-based code reuse and software-hardware testing feedback within a unified review process. Large language models and RL-based testing serve as auxiliary components for record organization, anomaly summarization, and simulation-based candidate scenario generation. Based on this setting, this paper analyzes preliminary evidence from multi-vehicle collaborative experimentation, code and experimental-skill sharing, and software-hardware collaborative testing. The analysis shows that experimental records can be examined together with their operating conditions, providing a reviewable basis for Apollo-on-Hongqi EV research.

cs.SE↗
arXiv2026

When Context Gets Root: Privilege Escalation in LLM Harnesses

Instruction hierarchy is a model-side defense that assigns instructions different levels of privilege according to their sources. These levels constrain which content may direct model behavior. During agent execution, however, agent harnesses construct context for each model invocation. This construction can elevate low-level content to a higher instruction level and grant it greater model-facing privilege. We introduce instruction privilege escalation. In this attack, an attacker induces an agent to elevate low-level malicious content to a higher instruction level. The elevated content then causes the agent to execute instructions it would not follow at their original level. We evaluate this threat by using multi-agent mechanisms to achieve 13 attack objectives across six coding-agent harnesses. These objectives span confidentiality, integrity, availability, and remote code execution. With unrestricted action execution, the attacks achieve all 13 objectives on all six harnesses. Under automatic permission review, the attacks achieve all 13 objectives on all three harnesses that provide this mode. We further reproduce the vulnerability using harness-provided persistent goals and scheduled tasks. These results demonstrate the generality of instruction privilege escalation.

cs.CR↗
arXiv2026

DSpark: Confidence-Scheduled Speculative Decoding with Semi-Autoregressive Generation

Speculative decoding accelerates Large Language Model (LLM) inference by decoupling draft generation from target verification. While recent parallel drafters efficiently propose long token sequences in a single forward pass, they suffer from rapid acceptance decay due to a lack of inter-token dependencies. Furthermore, indiscriminately verifying these extended blocks wastes critical batch capacity on tokens with high rejection risks, severely degrading throughput in high-concurrency serving systems. We introduce DSpark, a speculative decoding framework that unifies high-throughput parallel generation with adaptive, load-aware verification. To maintain draft quality, DSpark utilizes a semi-autoregressive architecture, coupling a parallel backbone with a lightweight sequential module, to introduce intra-block dependency modeling and mitigate suffix decay. To optimize system efficiency, DSpark employs confidence-scheduled verification, dynamically tailoring the verification length for each request based on estimated prefix survival probabilities and engine-specific throughput profiles. On offline benchmarks across diverse domains, DSpark substantially improves the accepted length over state-of-the-art autoregressive and parallel drafters. When deployed within the DeepSeek-V4 serving system under live user traffic, DSpark successfully mitigates verification waste. Compared to the established production baseline (MTP-1), DSpark accelerates per-user generation speeds by 60 to 85 percent at matched throughput levels. More importantly, by preventing severe throughput degradation under strict interactivity constraints, it enables performance tiers that were previously unattainable, shifting the Pareto frontier of our serving system.

cs.AI↗
arXiv2026

C2RM-Seg: Causal Counterfactual Reasoning with Structural-Semantic Priors for Weakly Supervised Histopathological Tissue Segmentation

Histopathological tissue segmentation is essential for computer-aided diagnosis, yet weakly supervised methods often suffer from noisy pseudo-labels generated by Class Activation Mapping (CAM). Existing CAM approaches tend to focus on staining-driven appearance cues rather than true causal tissue morphology, resulting in spurious localization and poor structural consistency. To address this issue, we propose C$^2$RM-Seg, a two-stage framework that integrates causal pseudo-label refinement with structure-aware semantic enhancement. For classification, we introduce a Causal Counterfactual Reasoning Module (C$^2$RM) that decomposes features into latent factors and performs counterfactual intervention via a learned causal structure matrix, suppressing confounding context and producing morphology-aligned CAMs. For segmentation, we design a Dual-Path Structural-Semantic Architecture that combines fine-grained structural features from ResNeSt with global semantic priors from a frozen DINOV3 foundation model. A cross-path gating mechanism adaptively regulates semantic injection using local structural cues to preserve boundary fidelity. To further mitigate residual pseudo-label noise, we propose an Uncertainty-Gated Margin (UGM) loss, which dynamically balances margin enforcement and confidence learning based on prediction uncertainty. Extensive experiments on two public histopathological tissue datasets show that C$^2$RM-Seg achieves state-of-the-art performance.

cs.CV↗
arXiv2026

PADD: Path-Aligned Decompression Distillation for Non-Router Teacher to Guide MoE Student Learning

As large language models (LLMs) continue to scale, it becomes increasingly challenging to grow model capacity under fixed computation budgets. We propose Path-Aligned Decompression Distillation (PADD), a framework for distilling knowledge from dense teachers without explicit routing into mixture-of-experts (MoE) students while learning high-quality routing policies. PADD organizes knowledge distillation into four stages in two phases: an initialization phase (Stage I) that builds diverse functionality in the student's experts through teacher neuron clustering and student-expert warmup, and a training phase (Stages II--IV) that integrates online adaptive distillation, path-refined policy optimization, and reward-augmented load balancing in a single training pipeline. Experiments on mathematical reasoning benchmarks demonstrate that PADD yields substantial gains over strong baselines at the same inference cost and that the MoE student can match or surpass its dense teacher. They also demonstrate effective teacher-to-student knowledge distillation and stable routing behavior.

cs.CL↗
arXiv2026

Discovering 100+ Compiler Defects in 72 Hours via LLM-Driven Semantic Logic Recomposition

Compilers constitute the foundational root-of-trust in software supply chains; however, their immense complexity inevitably conceals critical defects. Recent research has attempted to leverage historical bugs to design new mutation operators or fine-tune models to increase program diversity for compiler fuzzing.We observe, however, that bugs manifest primarily based on the semantics of input programs rather than their syntax. Unfortunately, current approaches, whether relying on syntactic mutation or general Large Language Model (LLM) fine-tuning, struggle to preserve the specific semantics found in the logic of bug-triggering programs. Consequently, these critical semantic triggers are often lost, resulting in a limitation of the diversity of generated programs. To explicitly reuse such semantics, we propose FeatureFuzz, a compiler fuzzer that combines features to generate programs. We define a feature as a decoupled primitive that encapsulates a natural language description of a bug-prone invariant, such as an out-of-bounds array access, alongside a concrete code witness of its realization. FeatureFuzz operates via a three-stage workflow: it first extracts features from historical bug reports, synthesizes coherent groups of features, and finally instantiates these groups into valid programs for compiler fuzzing. We evaluated FeatureFuzz on GCC and LLVM. Over 24-hour campaigns, FeatureFuzz uncovered 167 unique crashes, which is 2.78x more than the second-best fuzzer. Furthermore, through a 72-hour fuzzing campaign, FeatureFuzz identified 113 bugs in GCC and LLVM, 97 of which have already been confirmed by compiler developers, validating the approach's ability to stress-test modern compilers effectively.

cs.SE↗
arXiv2026

Mind the Gap: Action Rebinding Attacks against Android GUI Agents

Large multimodal model powered GUI agents are emerging as high-privilege operators on mobile platforms, entrusted to perceive screen content and inject inputs across application boundaries. While these agents aim to automate complex tasks, we demonstrate that their design introduces a fundamental conflict with Android's strict application sandboxing. We present a novel cross-application Action Rebinding attack, which allows a malicious application with zero dangerous permissions to hijack the agent's execution and perform privileged operations on behalf of the attacker. Our attack exploits the inevitable observation-action gap inherent in the agent's reasoning pipeline. A malicious app can render a benign ``contextual carrier'' to elicit a planned action, and then swap the foreground to a sensitive target application during the reasoning latency. The agent, unaware of the transition, unwittingly executes the action in the privileged context. We further advance this attack by weaponizing the agent's own task-recovery logic to create programmable, multi-step exploit loops , and introducing an Intent Alignment Strategy (IAS) that manipulates the agent's reasoning to rationalize the hijacked state. We evaluate our attack on six widely-used Android GUI agents. Our results demonstrate a 100% success rate for atomic action hijacking and the ability to orchestrate high-impact exploits, including unauthorized file deletion, SMS transmission, and app uninstallation, without the attacker holding any corresponding permissions. Furthermore, since the malicious application separates intent from capability and contains no privileged API calls, it achieves a 0% detection rate across commercial malware scanners (e.g., VirusTotal), highlighting a critical blind spot in current mobile security analysis. To access experimental logs and demonstration videos, please contact yi_qian@smail.nju.edu.cn.

cs.CR↗
arXiv2026

Tree Embedding in High Dimensions: Dynamic and Massively Parallel

Tree embedding has been a fundamental method in algorithm design with wide applications. We focus on the efficiency of building tree embedding in various computational settings under high-dimensional Euclidean $\mathbb{R}^d$. We devise a new tree embedding construction framework that operates on an arbitrary metric decomposition with bounded diameter, offering a tradeoff between distortion and the locality of its algorithmic steps. This framework works for general metric spaces and may be of independent interest beyond the Euclidean setting. Using this framework, we obtain a dynamic algorithm that maintains an $O_ε(\log n)$-distortion tree embedding with update time $\tilde O(n^ε+ d)$ subject to point insertions/deletions, and a massively parallel algorithm that achieves $O_ε(\log n)$-distortion in $O(1)$ rounds and total space $\tilde O(n^{1 + ε})$ (for constant $ε\in (0, 1)$). These new tree embedding results allow for a wide range of applications. Notably, under a similar performance guarantee as in our tree embedding algorithms, i.e., $\tilde O(n^ε+ d)$ update time and $O(1)$ rounds, we obtain $O_ε(\log n)$-approximate dynamic and MPC algorithms for $k$-median and earth-mover distance in $\mathbb{R}^d$.

cs.DS↗
arXiv2026

From Autoencoders to CycleGAN: Robust Unpaired Face Manipulation via Adversarial Learning

Human face synthesis and manipulation are increasingly important in entertainment and AI, with a growing demand for highly realistic, identity-preserving images even when only unpaired, unaligned datasets are available. We study unpaired face manipulation via adversarial learning, moving from autoencoder baselines to a robust, guided CycleGAN framework. While autoencoders capture coarse identity, they often miss fine details. Our approach integrates spectral normalization for stable training, identity- and perceptual-guided losses to preserve subject identity and high-level structure, and landmark-weighted cycle constraints to maintain facial geometry across pose and illumination changes. Experiments show that our adversarial trained CycleGAN improves realism (FID), perceptual quality (LPIPS), and identity preservation (ID-Sim) over autoencoders, with competitive cycle-reconstruction SSIM and practical inference times, which achieved high quality without paired datasets and approaching pix2pix on curated paired subsets. These results demonstrate that guided, spectrally normalized CycleGANs provide a practical path from autoencoders to robust unpaired face manipulation.

cs.LG↗
arXiv2025

Second-Order Fine-Tuning without Pain for LLMs:A Hessian Informed Zeroth-Order Optimizer

Fine-tuning large language models (LLMs) with classic first-order optimizers entails prohibitive GPU memory due to the backpropagation process. Recent works have turned to zeroth-order optimizers for fine-tuning, which save substantial memory by using two forward passes. However, these optimizers are plagued by the heterogeneity of parameter curvatures across different dimensions. In this work, we propose HiZOO, a diagonal Hessian informed zeroth-order optimizer which is the first work to leverage the diagonal Hessian to enhance zeroth-order optimizer for fine-tuning LLMs. What's more, HiZOO avoids the expensive memory cost and only increases one forward pass per step. Extensive experiments on various models (350M~66B parameters) indicate that HiZOO improves model convergence, significantly reducing training steps and effectively enhancing model accuracy. Moreover, we visualize the optimization trajectories of HiZOO on test functions, illustrating its effectiveness in handling heterogeneous curvatures. Lastly, we provide theoretical proofs of convergence for HiZOO. Code is publicly available at https://anonymous.4open.science/r/HiZOO27F8.

cs.LG↗
arXiv2024

Less is more: Embracing sparsity and interpolation with Esiformer for time series forecasting

Time series forecasting has played a significant role in many practical fields. But time series data generated from real-world applications always exhibits high variance and lots of noise, which makes it difficult to capture the inherent periodic patterns of the data, hurting the prediction accuracy significantly. To address this issue, we propose the Esiformer, which apply interpolation on the original data, decreasing the overall variance of the data and alleviating the influence of noise. What's more, we enhanced the vanilla transformer with a robust Sparse FFN. It can enhance the representation ability of the model effectively, and maintain the excellent robustness, avoiding the risk of overfitting compared with the vanilla implementation. Through evaluations on challenging real-world datasets, our method outperforms leading model PatchTST, reducing MSE by 6.5% and MAE by 5.8% in multivariate time series forecasting. Code is available at: https://github.com/yyg1282142265/Esiformer/tree/main.

cs.LG↗
arXiv2024

Backdoor Attacks and Defenses on Semantic-Symbol Reconstruction in Semantic Communications

Semantic communication is of crucial importance for the next-generation wireless communication networks. The existing works have developed semantic communication frameworks based on deep learning. However, systems powered by deep learning are vulnerable to threats such as backdoor attacks and adversarial attacks. This paper delves into backdoor attacks targeting deep learning-enabled semantic communication systems. Since current works on backdoor attacks are not tailored for semantic communication scenarios, a new backdoor attack paradigm on semantic symbols (BASS) is introduced, based on which the corresponding defense measures are designed. Specifically, a training framework is proposed to prevent BASS. Additionally, reverse engineering-based and pruning-based defense strategies are designed to protect against backdoor attacks in semantic communication. Simulation results demonstrate the effectiveness of both the proposed attack paradigm and the defense strategies.

cs.CR↗
arXiv2024

Approximate Wireless Communication for Lossy Gradient Updates in IoT Federated Learning

Federated learning (FL) has emerged as a distributed machine learning (ML) technique that can protect local data privacy for participating clients and improve system efficiency. Instead of sharing raw data, FL exchanges intermediate learning parameters, such as gradients, among clients. This article presents an efficient wireless communication approach tailored for FL parameter transmission, especially for Internet of Things (IoT) devices, to facilitate model aggregation. Our study considers practical wireless channels that can lead to random bit errors, which can substantially affect FL performance. Motivated by empirical gradient value distribution, we introduce a novel received bit masking method that confines received gradient values within prescribed limits. Moreover, given the intrinsic error resilience of ML gradients, our approach enables the delivery of approximate gradient values with errors without resorting to extensive error correction coding or retransmission. This strategy reduces computational overhead at both the transmitter and the receiver and minimizes communication latency. Consequently, our scheme is particularly well-suited for resource-constrained IoT devices. Additionally, we explore the inherent protection of the most significant bits (MSBs) through gray coding in high-order modulation. Our simulations demonstrate that our proposed scheme can effectively mitigate random bit errors in FL performance, achieving similar learning objectives, but with the 50% air time required by existing methods involving error correction and retransmission.

cs.IT↗
arXiv2024

Sparse-VQ Transformer: An FFN-Free Framework with Vector Quantization for Enhanced Time Series Forecasting

Time series analysis is vital for numerous applications, and transformers have become increasingly prominent in this domain. Leading methods customize the transformer architecture from NLP and CV, utilizing a patching technique to convert continuous signals into segments. Yet, time series data are uniquely challenging due to significant distribution shifts and intrinsic noise levels. To address these two challenges,we introduce the Sparse Vector Quantized FFN-Free Transformer (Sparse-VQ). Our methodology capitalizes on a sparse vector quantization technique coupled with Reverse Instance Normalization (RevIN) to reduce noise impact and capture sufficient statistics for forecasting, serving as an alternative to the Feed-Forward layer (FFN) in the transformer architecture. Our FFN-free approach trims the parameter count, enhancing computational efficiency and reducing overfitting. Through evaluations across ten benchmark datasets, including the newly introduced CAISO dataset, Sparse-VQ surpasses leading models with a 7.84% and 4.17% decrease in MAE for univariate and multivariate time series forecasting, respectively. Moreover, it can be seamlessly integrated with existing transformer-based models to elevate their performance.

cs.LG↗
arXiv2023

GCformer: An Efficient Framework for Accurate and Scalable Long-Term Multivariate Time Series Forecasting

Transformer-based models have emerged as promising tools for time series forecasting. However, these model cannot make accurate prediction for long input time series. On the one hand, they failed to capture global dependencies within time series data. On the other hand, the long input sequence usually leads to large model size and high time complexity. To address these limitations, we present GCformer, which combines a structured global convolutional branch for processing long input sequences with a local Transformer-based branch for capturing short, recent signals. A cohesive framework for a global convolution kernel has been introduced, utilizing three distinct parameterization methods. The selected structured convolutional kernel in the global branch has been specifically crafted with sublinear complexity, thereby allowing for the efficient and effective processing of lengthy and noisy input signals. Empirical studies on six benchmark datasets demonstrate that GCformer outperforms state-of-the-art methods, reducing MSE error in multivariate time series benchmarks by 4.38% and model parameters by 61.92%. In particular, the global convolutional branch can serve as a plug-in block to enhance the performance of other models, with an average improvement of 31.93\%, including various recently published Transformer-based models. Our code is publicly available at https://github.com/zyj-111/GCformer.

cs.LG↗
arXiv2023

CSMAAFL: Client Scheduling and Model Aggregation in Asynchronous Federated Learning

Asynchronous federated learning aims to solve the straggler problem in heterogeneous environments, i.e., clients have small computational capacities that could cause aggregation delay. The principle of asynchronous federated learning is to allow the server to aggregate the model once it receives an update from any client rather than waiting for updates from multiple clients or waiting a specified amount of time in the synchronous mode. Due to the asynchronous setting, the stale model problem could occur, where the slow clients could utilize an outdated local model for their local data training. Consequently, when these locally trained models are uploaded to the server, they may impede the convergence of the global training. Therefore, effective model aggregation strategies play a significant role in updating the global model. Besides, client scheduling is also critical when heterogeneous clients with diversified computing capacities are participating in the federated learning process. This work first investigates the impact of the convergence of asynchronous federated learning mode when adopting the aggregation coefficient in synchronous mode. The effective aggregation solutions that can achieve the same convergence result as in the synchronous mode are then proposed, followed by an improved aggregation method with client scheduling. The simulation results in various scenarios demonstrate that the proposed algorithm converges with a similar level of accuracy as the classical synchronous federated learning algorithm but effectively accelerates the learning process, especially in its early stage.

cs.DC↗
arXiv2023

Approximate Wireless Communication for Federated Learning

This paper presents an approximate wireless communication scheme for federated learning (FL) model aggregation in the uplink transmission. We consider a realistic channel that reveals bit errors during FL model exchange in wireless networks. Our study demonstrates that random bit errors during model transmission can significantly affect FL performance. To overcome this challenge, we propose an approximate communication scheme based on the mathematical and statistical proof that machine learning (ML) model gradients are bounded under certain constraints. This bound enables us to introduce a novel encoding scheme for float-to-binary representation of gradient values and their QAM constellation mapping. Besides, since FL gradients are error-resilient, the proposed scheme simply delivers gradients with errors when the channel quality is satisfactory, eliminating extensive error-correcting codes and/or retransmission. The direct benefits include less overhead and lower latency. The proposed scheme is well-suited for resource-constrained devices in wireless networks. Through simulations, we show that the proposed scheme is effective in reducing the impact of bit errors on FL performance and saves at least half the time than transmission with error correction and retransmission to achieve the same learning performance. In addition, we investigated the effectiveness of bit protection mechanisms in high-order modulation when gray coding is employed and found that this approach considerably enhances learning performance.

cs.DC↗
arXiv2022

Silicon-based decoder for polarization-encoding quantum key distribution

Silicon-based polarization-encoding quantum key distribution (QKD) has been widely studied, owing to its low cost and robustness. However, prior studies have utilized off-chip devices to demodulate the quantum states or perform polarization compensation, given the difficulty of fabricating polarized independent components on the chip. In this paper, we propose a fully chip-based decoder for polarization-encoding QKD. The chip realizes a polarization state analyzer and compensates for the BB84 protocol without requiring additional hardware. It is based on a polarization-to-path conversion method that uses a polarization splitter-rotator. The chip was fabricated using a standard silicon photonics foundry; it has a compact design and is suitable for mass production. In the experimental stability test, an average quantum bit error rate of $0.56\%$ was achieved through continuous operation for 10 h without any polarization feedback. Furthermore, using the developed feedback algorithm, the chip enabled the automatic compensation of the fiber polarization drift, which was emulated by a random fiber polarization scrambler. In the case of the QKD demonstration, we obtained a finite-key secret rate of 240 bps over a fiber spool of 100 km. This study represents an important step toward the integrated, practical, and large-scale deployment of QKD systems.

quant-ph↗