arXiv subjects

Yuhao Zhang

Publications and source records attributed to Yuhao Zhang.

At least 109 records · Page 6Linked to original sources

arXiv2024

Robust Model Predictive Control for Constrained Uncertain Systems Based on Concentric Container and Varying Tube

This paper proposes a novel robust model predictive control (RMPC) method for the stabilization of constrained systems subject to additive disturbance (AD) and multiplicative disturbance (MD). Concentric containers are introduced to facilitate the characterization of MD, and varying tubes are constructed to bound reachable states. By restricting states and the corresponding inputs in containers with free sizes and a fixed shape, feasible MDs, which are the products of model uncertainty with states and inputs, are restricted into polytopes with free sizes. Then, tubes with different centers and shapes are constructed based on the nominal dynamics and the knowledge of AD and MD. The free sizes of containers allow for a more accurate characterization of MD, while the fixed shape reduces online computational burden, making the proposed method less conservative and computationally efficient. Moreover, the shape of containers is optimized to further reduce conservativeness. Compared to the RMPC method using homothetic tubes, the proposed method has a larger region of attraction while involving fewer decision variables and constraints in the online optimization problem.

eess.SY↗

arXiv2024

Edge Arrival Online Matching: The Power of Free Disposal on Acyclic Graphs

Online matching is a fundamental problem in the study of online algorithms. We study the problem under a very general arrival model: the edge arrival model. Free disposal is an important notion in the online matching literature, which allows the algorithm to dispose of the previously matched edges. Without free disposal, we cannot achieve any bounded ratio, even with randomized algorithms, when edges are weighted. Our paper focuses on clarifying the power of free disposal in both the unweighted and the weighted setting. As far as we know, it's still uncertain if free disposal can give us extra leverage to enhance the competitive ratio in the unweighted scenario, even in specific instances such as Growing Trees, where every new edge adds a new leaf to the graph. Our study serves as a valuable initial exploration of this open question. The results are listed as follows: 1. With free disposal, we improve the competitive ratio for unweighted online matching on Growing Trees from $5/9$ to $2/3 \approx 0.66$, and show that the ratio is tight. For Forests, a more general setting where the underlying graph is a forest and edges may arrive in arbitrary order, we improve the competitive ratio from $5/9$ to $5/8 = 0.625$. 2. Both the ratios of $2/3$ and $0.625$ show a separation to the upper bound of the competitive ratio without free disposal on Growing Trees ($0.5914$). Therefore, we demonstrate the additional power of free disposal for the unweighted setting for the first time, at least in the special setting of Growing Trees and Forests. 3. We improve the competitive ratio for weighted online matching on Growing Trees from $1/3$ to $1/2$ using a very simple ordinal algorithm, and show that it is optimal among ordinal algorithms.

cs.DS↗

arXiv2024

CodeFort: Robust Training for Code Generation Models

Code generation models are not robust to small perturbations, which often lead to incorrect generations and significantly degrade the performance of these models. Although improving the robustness of code generation models is crucial to enhancing user experience in real-world applications, existing research efforts do not address this issue. To fill this gap, we propose CodeFort, a framework to improve the robustness of code generation models, generalizing a large variety of code perturbations to enrich the training data and enabling various robust training strategies, mixing data augmentation, batch augmentation, adversarial logits pairing, and contrastive learning, all carefully designed to support high-throughput training. Extensive evaluations show that we increase the average robust pass rates of baseline CodeGen models from 14.79 to 21.74. We notably decrease the robustness drop rate from 95.02% to 54.95% against code-syntax perturbations.

cs.SE↗

arXiv2024

Roadmap towards Superhuman Speech Understanding using Large Language Models

The success of large language models (LLMs) has prompted efforts to integrate speech and audio data, aiming to create general foundation models capable of processing both textual and non-textual inputs. Recent advances, such as GPT-4o, highlight the potential for end-to-end speech LLMs, which preserves non-semantic information and world knowledge for deeper speech understanding. To guide the development of speech LLMs, we propose a five-level roadmap, ranging from basic automatic speech recognition (ASR) to advanced superhuman models capable of integrating non-semantic information with abstract acoustic knowledge for complex tasks. Moreover, we design a benchmark, SAGI Bechmark, that standardizes critical aspects across various tasks in these five levels, uncovering challenges in using abstract acoustic knowledge and completeness of capability. Our findings reveal gaps in handling paralinguistic cues and abstract acoustic knowledge, and we offer future directions. This paper outlines a roadmap for advancing speech LLMs, introduces a benchmark for evaluation, and provides key insights into their current limitations and potential.

cs.CL↗

arXiv2024

A Framework for Reproducible Benchmarking and Performance Diagnosis of SLAM Systems

We propose SLAMFuse, an open-source SLAM benchmarking framework that provides consistent crossplatform environments for evaluating multi-modal SLAM algorithms, along with tools for data fuzzing, failure detection, and diagnosis across different datasets. Our framework introduces a fuzzing mechanism to test the resilience of SLAM algorithms against dataset perturbations. This enables the assessment of pose estimation accuracy under varying conditions and identifies critical perturbation thresholds. SLAMFuse improves diagnostics with failure detection and analysis tools, examining algorithm behaviour against dataset characteristics. SLAMFuse uses Docker to ensure reproducible testing conditions across diverse datasets and systems by streamlining dependency management. Emphasizing the importance of reproducibility and introducing advanced tools for algorithm evaluation and performance diagnosis, our work sets a new precedent for reliable benchmarking of SLAM systems. We provide ready-to-use docker compatible versions of the algorithms and datasets used in the experiments, together with guidelines for integrating and benchmarking new algorithms. Code is available at https://github.com/nikolaradulov/slamfuse

cs.RO↗

arXiv2024

RAG-QA Arena: Evaluating Domain Robustness for Long-form Retrieval Augmented Question Answering

Question answering based on retrieval augmented generation (RAG-QA) is an important research topic in NLP and has a wide range of real-world applications. However, most existing datasets for this task are either constructed using a single source corpus or consist of short extractive answers, which fall short of evaluating large language model (LLM) based RAG-QA systems on cross-domain generalization. To address these limitations, we create Long-form RobustQA (LFRQA), a new dataset comprising human-written long-form answers that integrate short extractive answers from multiple documents into a single, coherent narrative, covering 26K queries and large corpora across seven different domains. We further propose RAG-QA Arena by directly comparing model-generated answers against LFRQA's answers using LLMs as evaluators. We show via extensive experiments that RAG-QA Arena and human judgments on answer quality are highly correlated. Moreover, only 41.3% of the most competitive LLM's answers are preferred to LFRQA's answers, demonstrating RAG-QA Arena as a challenging evaluation platform for future research.

cs.CL↗

arXiv2024

Promptriever: Instruction-Trained Retrievers Can Be Prompted Like Language Models

Instruction-tuned language models (LM) are able to respond to imperative commands, providing a more natural user interface compared to their base counterparts. In this work, we present Promptriever, the first retrieval model able to be prompted like an LM. To train Promptriever, we curate and release a new instance-level instruction training set from MS MARCO, spanning nearly 500k instances. Promptriever not only achieves strong performance on standard retrieval tasks, but also follows instructions. We observe: (1) large gains (reaching SoTA) on following detailed relevance instructions (+14.3 p-MRR / +3.1 nDCG on FollowIR), (2) significantly increased robustness to lexical choices/phrasing in the query+instruction (+12.9 Robustness@10 on InstructIR), and (3) the ability to perform hyperparameter search via prompting to reliably improve retrieval performance (+1.4 average increase on BEIR). Promptriever demonstrates that retrieval models can be controlled with prompts on a per-query basis, setting the stage for future work aligning LM prompting techniques with information retrieval.

cs.IR↗

arXiv2024

Almost Tight Approximation Hardness for Single-Source Directed k-Edge-Connectivity

In the $k$-connected directed Steiner tree problem ($k$-DST), we are given an $n$-vertex directed graph $G=(V,E)$ with edge costs, a connectivity requirement $k$, a root $r\in V$ and a set of terminals $T\subseteq V$. The goal is to find a minimum-cost subgraph $H\subseteq G$ that has $k$ internally disjoint paths from the root vertex $r$ to every terminal $t\in T$. In this paper, we show the approximation hardness of $k$-DST for various parameters, which thus close some long-standing open problems. - $Ω\left(|T|/\log |T|\right)$-approximation hardness, which holds under the standard assumption $\mathrm{NP}\neq \mathrm{ZPP}$. The inapproximability ratio is tightened to $Ω\left(|T|\right)$ under the Strongish Planted Clique Hypothesis [Manurangsi, Rubinstein and Schramm, ITCS 2021]. The latter hardness result matches the approximation ratio of $|T|$ obtained by a trivial approximation algorithm, thus closing the long-standing open problem. - $Ω\left(\sqrt{2}^k / k\right)$-approximation hardness for the general case of $k$-DST under the assumption $\mathrm{NP}\neq\mathrm{ZPP}$. This is the first hardness result known for survivable network design problems with an inapproximability ratio exponential in $k$. - $Ω\left((k/L)^{L/4}\right)$-approximation hardness for $k$-DST on $L$-layered graphs for $L\le O\left(\log n\right)$. This almost matches the approximation ratio of $O(k^{L-1}\cdot L \cdot \log |T|)$ achieving in $O\left(n^L\right)$-time due to Laekhanukit [ICALP`16].

cs.DS↗

arXiv2024

Dancing in Chains: Reconciling Instruction Following and Faithfulness in Language Models

Modern language models (LMs) need to follow human instructions while being faithful; yet, they often fail to achieve both. Here, we provide concrete evidence of a trade-off between instruction following (i.e., follow open-ended instructions) and faithfulness (i.e., ground responses in given context) when training LMs with these objectives. For instance, fine-tuning LLaMA-7B on instruction following datasets renders it less faithful. Conversely, instruction-tuned Vicuna-7B shows degraded performance at following instructions when further optimized on tasks that require contextual grounding. One common remedy is multi-task learning (MTL) with data mixing, yet it remains far from achieving a synergic outcome. We propose a simple yet effective method that relies on Rejection Sampling for Continued Self-instruction Tuning (ReSet), which significantly outperforms vanilla MTL. Surprisingly, we find that less is more, as training ReSet with high-quality, yet substantially smaller data (three-fold less) yields superior results. Our findings offer a better understanding of objective discrepancies in alignment training of LMs.

cs.CL↗

arXiv2024

DragVideo: Interactive Drag-style Video Editing

Video generation models have shown their superior ability to generate photo-realistic video. However, how to accurately control (or edit) the video remains a formidable challenge. The main issues are: 1) how to perform direct and accurate user control in editing; 2) how to execute editings like changing shape, expression, and layout without unsightly distortion and artifacts to the edited content; and 3) how to maintain spatio-temporal consistency of video after editing. To address the above issues, we propose DragVideo, a general drag-style video editing framework. Inspired by DragGAN, DragVideo addresses issues 1) and 2) by proposing the drag-style video latent optimization method which gives desired control by updating noisy video latent according to drag instructions through video-level drag objective function. We amend issue 3) by integrating the video diffusion model with sample-specific LoRA and Mutual Self-Attention in DragVideo to ensure the edited result is spatio-temporally consistent. We also present a series of testing examples for drag-style video editing and conduct extensive experiments across a wide array of challenging editing tasks, such as motion, skeleton editing, etc, underscoring DragVideo can edit video in an intuitive, faithful to the user's intention manner, with nearly unnoticeable distortion and artifacts, while maintaining spatio-temporal consistency. While traditional prompt-based video editing fails to do the former two and directly applying image drag editing fails in the last, DragVideo's versatility and generality are emphasized. Github link: https://github.com/RickySkywalker/DragVideo-Official.

cs.GR↗

arXiv2024

A One-Layer Decoder-Only Transformer is a Two-Layer RNN: With an Application to Certified Robustness

This paper reveals a key insight that a one-layer decoder-only Transformer is equivalent to a two-layer Recurrent Neural Network (RNN). Building on this insight, we propose ARC-Tran, a novel approach for verifying the robustness of decoder-only Transformers against arbitrary perturbation spaces. Compared to ARC-Tran, current robustness verification techniques are limited either to specific and length-preserving perturbations like word substitutions or to recursive models like LSTMs. ARC-Tran addresses these limitations by meticulously managing position encoding to prevent mismatches and by utilizing our key insight to achieve precise and scalable verification. Our evaluation shows that ARC-Tran (1) trains models more robust to arbitrary perturbation spaces than those produced by existing techniques and (2) shows high certification accuracy of the resulting models.

cs.CL↗

arXiv2024

PECAN: A Deterministic Certified Defense Against Backdoor Attacks

Neural networks are vulnerable to backdoor poisoning attacks, where the attackers maliciously poison the training set and insert triggers into the test input to change the prediction of the victim model. Existing defenses for backdoor attacks either provide no formal guarantees or come with expensive-to-compute and ineffective probabilistic guarantees. We present PECAN, an efficient and certified approach for defending against backdoor attacks. The key insight powering PECAN is to apply off-the-shelf test-time evasion certification techniques on a set of neural networks trained on disjoint partitions of the data. We evaluate PECAN on image classification and malware detection datasets. Our results demonstrate that PECAN can (1) significantly outperform the state-of-the-art certified backdoor defense, both in defense strength and efficiency, and (2) on real back-door attacks, PECAN can reduce attack success rate by order of magnitude when compared to a range of baselines from the literature.

cs.CR↗

arXiv2024

Ensuring Safe and High-Quality Outputs: A Guideline Library Approach for Language Models

Large Language Models (LLMs) exhibit impressive capabilities but also present risks such as biased content generation and privacy issues. One of the current alignment techniques includes principle-driven integration, but it faces challenges arising from the imprecision of manually crafted rules and inadequate risk perception in models without safety training. To address these, we introduce Guide-Align, a two-stage approach. Initially, a safety-trained model identifies potential risks and formulates specific guidelines for various inputs, establishing a comprehensive library of guidelines and a model for input-guidelines retrieval. Subsequently, the retrieval model correlates new inputs with relevant guidelines, which guide LLMs in response generation to ensure safe and high-quality outputs, thereby aligning with human values. An additional optional stage involves fine-tuning a model with well-aligned datasets generated through the process implemented in the second stage. Our method customizes guidelines to accommodate diverse inputs, thereby enhancing the fine-grainedness and comprehensiveness of the guideline library. Furthermore, it incorporates safety expertise from a safety-trained LLM through a lightweight retrieval model. We evaluate our approach on three benchmarks, demonstrating significant improvements in LLM security and quality. Notably, our fine-tuned model, Labrador, even at 13 billion parameters, outperforms GPT-3.5-turbo and surpasses GPT-4 in alignment capabilities.

cs.CL↗

arXiv2024

Hybrid Zonotope-Based Backward Reachability Analysis for Neural Feedback Systems With Nonlinear Plant Models

The increasing prevalence of neural networks in safety-critical control systems underscores the imperative need for rigorous methods to ensure the reliability and safety of these systems. This work introduces a novel approach employing hybrid zonotopes to compute the over-approximation of backward reachable sets for neural feedback systems with nonlinear plant models and general activation functions. Closed-form expressions as hybrid zonotopes are provided for the over-approximated backward reachable sets, and a refinement procedure is proposed to alleviate the potential conservatism of the approximation. Two numerical examples are provided to illustrate the effectiveness of the proposed approach.

math.OC↗

arXiv2024

Towards Lossless ANN-SNN Conversion under Ultra-Low Latency with Dual-Phase Optimization

Spiking neural networks (SNNs) operating with asynchronous discrete events show higher energy efficiency with sparse computation. A popular approach for implementing deep SNNs is ANN-SNN conversion combining both efficient training of ANNs and efficient inference of SNNs. However, the accuracy loss is usually non-negligible, especially under a few time steps, which restricts the applications of SNN on latency-sensitive edge devices greatly. In this paper, we first identify that such performance degradation stems from the misrepresentation of the negative or overflow residual membrane potential in SNNs. Inspired by this, we decompose the conversion error into three parts: quantization error, clipping error, and residual membrane potential representation error. With such insights, we propose a two-stage conversion algorithm to minimize those errors respectively. Besides, We show each stage achieves significant performance gains in a complementary manner. By evaluating on challenging datasets including CIFAR-10, CIFAR- 100 and ImageNet, the proposed method demonstrates the state-of-the-art performance in terms of accuracy, latency and energy preservation. Furthermore, our method is evaluated using a more challenging object detection task, revealing notable gains in regression performance under ultra-low latency when compared to existing spike-based detection algorithms. Codes are available at https://github.com/Windere/snn-cvt-dual-phase.

cs.NE↗

arXiv2024

Verified Training for Counterfactual Explanation Robustness under Data Shift

Counterfactual explanations (CEs) enhance the interpretability of machine learning models by describing what changes to an input are necessary to change its prediction to a desired class. These explanations are commonly used to guide users' actions, e.g., by describing how a user whose loan application was denied can be approved for a loan in the future. Existing approaches generate CEs by focusing on a single, fixed model, and do not provide any formal guarantees on the CEs' future validity. When models are updated periodically to account for data shift, if the generated CEs are not robust to the shifts, users' actions may no longer have the desired impacts on their predictions. This paper introduces VeriTraCER, an approach that jointly trains a classifier and an explainer to explicitly consider the robustness of the generated CEs to small model shifts. VeriTraCER optimizes over a carefully designed loss function that ensures the verifiable robustness of CEs to local model updates, thus providing deterministic guarantees to CE validity. Our empirical evaluation demonstrates that VeriTraCER generates CEs that (1) are verifiably robust to small model updates and (2) display competitive robustness to state-of-the-art approaches in handling empirical model updates including random initialization, leave-one-out, and distribution shifts.

cs.LG↗

arXiv2023

Soft Alignment of Modality Space for End-to-end Speech Translation

End-to-end Speech Translation (ST) aims to convert speech into target text within a unified model. The inherent differences between speech and text modalities often impede effective cross-modal and cross-lingual transfer. Existing methods typically employ hard alignment (H-Align) of individual speech and text segments, which can degrade textual representations. To address this, we introduce Soft Alignment (S-Align), using adversarial training to align the representation spaces of both modalities. S-Align creates a modality-invariant space while preserving individual modality quality. Experiments on three languages from the MuST-C dataset show S-Align outperforms H-Align across multiple tasks and offers translation capabilities on par with specialized translation models.

cs.CL↗

arXiv2023

Channel sensing for holographic interference surfaces based on the principle of interferometry

The Holographic Interference Surface (HIS) provides a new paradigm for building a more cost-effective wireless communication architecture. In this paper, we derive the principles of holographic interference theory for electromagnetic wave reception and transmission, whereby the optical holography is extended to communication holography and a channel sensing architecture for holographic interference surfaces is established. Unlike the traditional pilot-based channel estimation approaches, the proposed architecture circumvents the complicated processes like filtering, analog to digital conversion (ADC), down conversion. Instead, it relies on interfering the object waves with a pre-designed reference wave, and therefore reduces the hardware complexity and requires less time-frequency resources for channel estimation. To address the self-interference problem in the holographic recording process, we propose a phase shifting-based interference suppression (PSIS) method according to the structural characteristics of communication hologram and interference composition. We then propose a Prony-based multi-user channel segmentation (PMCS) algorithm to acquire the channel state information (CSI). Our theoretical analysis shows that the estimation error of the PMCS algorithm converges to zero when the number of HIS units is large enough. Simulation results show that under the holographic architecture, our proposed algorithm can accurately estimate the CSI in multi-user scenarios.

eess.SP↗