arXiv subjects

Jun Sun

Publications and source records attributed to Jun Sun.

At least 127 records · Page 7Linked to original sources

arXiv2025

FixDrive: Automatically Repairing Autonomous Vehicle Driving Behaviour for $0.08 per Violation

Autonomous Vehicles (AVs) are advancing rapidly, with Level-4 AVs already operating in real-world conditions. Current AVs, however, still lag behind human drivers in adaptability and performance, often exhibiting overly conservative behaviours and occasionally violating traffic laws. Existing solutions, such as runtime enforcement, mitigate this by automatically repairing the AV's planned trajectory at runtime, but such approaches lack transparency and should be a measure of last resort. It would be preferable for AV repairs to generalise beyond specific incidents and to be interpretable for users. In this work, we propose FixDrive, a framework that analyses driving records from near-misses or law violations to generate AV driving strategy repairs that reduce the chance of such incidents occurring again. These repairs are captured in μDrive, a high-level domain-specific language for specifying driving behaviours in response to event-based triggers. Implemented for the state-of-the-art autonomous driving system Apollo, FixDrive identifies and visualises critical moments from driving records, then uses a Multimodal Large Language Model (MLLM) with zero-shot learning to generate μDrive programs. We tested FixDrive on various benchmark scenarios, and found that the generated repairs improved the AV's performance with respect to following traffic laws, avoiding collisions, and successfully reaching destinations. Furthermore, the direct costs of repairing an AV -- 15 minutes of offline analysis and $0.08 per violation -- are reasonable in practice.

cs.SE↗

arXiv2025

Democratic Training Against Universal Adversarial Perturbations

Despite their advances and success, real-world deep neural networks are known to be vulnerable to adversarial attacks. Universal adversarial perturbation, an input-agnostic attack, poses a serious threat for them to be deployed in security-sensitive systems. In this case, a single universal adversarial perturbation deceives the model on a range of clean inputs without requiring input-specific optimization, which makes it particularly threatening. In this work, we observe that universal adversarial perturbations usually lead to abnormal entropy spectrum in hidden layers, which suggests that the prediction is dominated by a small number of ``feature'' in such cases (rather than democratically by many features). Inspired by this, we propose an efficient yet effective defense method for mitigating UAPs called \emph{Democratic Training} by performing entropy-based model enhancement to suppress the effect of the universal adversarial perturbations in a given model. \emph{Democratic Training} is evaluated with 7 neural networks trained on 5 benchmark datasets and 5 types of state-of-the-art universal adversarial attack methods. The results show that it effectively reduces the attack success rate, improves model robustness and preserves the model accuracy on clean samples.

cs.LG↗

arXiv2024

Training Verification-Friendly Neural Networks via Neuron Behavior Consistency

Formal verification provides critical security assurances for neural networks, yet its practical application suffers from the long verification time. This work introduces a novel method for training verification-friendly neural networks, which are robust, easy to verify, and relatively accurate. Our method integrates neuron behavior consistency into the training process, making neuron activation states remain consistent across different inputs within a local neighborhood. This reduces the number of unstable neurons and tightens the bounds of neurons thereby enhancing the network's verifiability. We evaluated our method using the MNIST, Fashion-MNIST, and CIFAR-10 datasets with various network architectures. The experimental results demonstrate that networks trained using our method are verification-friendly across different radii and architectures, whereas other tools fail to maintain verifiability as the radius increases. Additionally, we show that our method can be combined with existing approaches to further improve the verifiability of networks.

cs.LG↗

arXiv2024

PonziLens+: Visualizing Bytecode Actions for Smart Ponzi Scheme Identification

With the prevalence of smart contracts, smart Ponzi schemes have become a common fraud on blockchain and have caused significant financial loss to cryptocurrency investors in the past few years. Despite the critical importance of detecting smart Ponzi schemes, a reliable and transparent identification approach adaptive to various smart Ponzi schemes is still missing. To fill the research gap, we first extract semantic-meaningful actions to represent the execution behaviors specified in smart contract bytecodes, which are derived from a literature review and in-depth interviews with domain experts. We then propose PonziLens+, a novel visual analytic approach that provides an intuitive and reliable analysis of Ponzi-scheme-related features within these execution behaviors. PonziLens+ has three visualization modules that intuitively reveal all potential behaviors of a smart contract, highlighting fraudulent features across three levels of detail. It can help smart contract investors and auditors achieve confident identification of any smart Ponzi schemes. We conducted two case studies and in-depth user interviews with 12 domain experts and common investors to evaluate PonziLens+. The results demonstrate the effectiveness and usability of PonziLens+ in achieving an effective identification of smart Ponzi schemes.

cs.HC↗

arXiv2024

PrettiSmart: Visual Interpretation of Smart Contracts via Simulation

Smart contracts are the fundamental components of blockchain technology. They are programs to determine cryptocurrency transactions, and are irreversible once deployed, making it crucial for cryptocurrency investors to understand the cryptocurrency transaction behaviors of smart contracts comprehensively. However, it is a challenging (if not impossible) task for investors, as they do not necessarily have a programming background to check the complex source code. Even for investors with certain programming skills, inferring all the potential behaviors from the code alone is still difficult, since the actual behaviors can be different when different investors are involved. To address this challenge, we propose PrettiSmart, a novel visualization approach via execution simulation to achieve intuitive and reliable visual interpretation of smart contracts. Specifically, we develop a simulator to comprehensively capture most of the possible real-world smart contract behaviors, involving multiple investors and various smart contract functions. Then, we present PrettiSmart to intuitively visualize the simulation results of a smart contract, which consists of two modules: The Simulation Overview Module is a barcode-based design, providing a visual summary for each simulation, and the Simulation Detail Module is an augmented sequential design to display the cryptocurrency transaction details in each simulation, such as function call sequences, cryptocurrency flows, and state variable changes. It can allow investors to intuitively inspect and understand how a smart contract will work. We evaluate PrettiSmart through two case studies and in-depth user interviews with 12 investors. The results demonstrate the effectiveness and usability of PrettiSmart in facilitating an easy interpretation of smart contracts.

cs.HC↗

arXiv2024

Do Influence Functions Work on Large Language Models?

Influence functions are important for quantifying the impact of individual training data points on a model's predictions. Although extensive research has been conducted on influence functions in traditional machine learning models, their application to large language models (LLMs) has been limited. In this work, we conduct a systematic study to address a key question: do influence functions work on LLMs? Specifically, we evaluate influence functions across multiple tasks and find that they consistently perform poorly in most settings. Our further investigation reveals that their poor performance can be attributed to: (1) inevitable approximation errors when estimating the iHVP component due to the scale of LLMs, (2) uncertain convergence during fine-tuning, and, more fundamentally, (3) the definition itself, as changes in model parameters do not necessarily correlate with changes in LLM behavior. Thus, our study suggests the need for alternative approaches for identifying influential samples.

cs.CL↗

arXiv2024

Unleashing the Unseen: Harnessing Benign Datasets for Jailbreaking Large Language Models

Despite significant ongoing efforts in safety alignment, large language models (LLMs) such as GPT-4 and LLaMA 3 remain vulnerable to jailbreak attacks that can induce harmful behaviors, including through the use of adversarial suffixes. Building on prior research, we hypothesize that these adversarial suffixes are not mere bugs but may represent features that can dominate the LLM's behavior. To evaluate this hypothesis, we conduct several experiments. First, we demonstrate that benign features can be effectively made to function as adversarial suffixes, i.e., we develop a feature extraction method to extract sample-agnostic features from benign dataset in the form of suffixes and show that these suffixes may effectively compromise safety alignment. Second, we show that adversarial suffixes generated from jailbreak attacks may contain meaningful features, i.e., appending the same suffix to different prompts results in responses exhibiting specific characteristics. Third, we show that such benign-yet-safety-compromising features can be easily introduced through fine-tuning using only benign datasets. As a result, we are able to completely eliminate GPT's safety alignment in a blackbox setting through finetuning with only benign data. Our code and data is available at \url{https://github.com/suffix-maybe-feature/adver-suffix-maybe-features}.

cs.CR↗

arXiv2024

Domain-Pair Intertwined Topological Domain Structure in Elemental Bi Monolayer

Ferroelectric domain structures, separated by domain walls, often display unconventional physics and hold significant potential for applications in nano-devices. Most naturally growth domain walls are charge-neutral to avoid increased electrostatic energy, while the intrinsically stable charged 180° domain walls in Bi monolayer challenged this conventional knowledge and emerged an unexplored field. Here, using machine-learning potential and molecular dynamics (MD) simulations, we investigated the finite-temperature dynamics of domain walls and discovered a domain-pair intertwined topological domain structure in Bi monolayer. In 180° domain walls, a unique polarization switching mechanism is observed, characterized by the out-of-plane shuffle of Bi atoms without bond breaking. This shuffle mechanism reverses the charge properties of Bi atoms, transforming Bi anions into cations and vice versa, ultimately reversing the polarization. Then, we observed a topological multi-domain structure with two groups of domain pairs intertwined. The charged 180° domain walls form local domain pairs, with the 90° domain walls emerge between different domain pairs. This multi-domain maintains a stable topological structure within the strain range (ε_x = 0 to 4.70%) and exhibits rich domain wall reactions under further applied strain. Our findings provide insights into the charged 180° domain walls and the related topological domain structures, enabling new opportunities for applications in electronic and nano-electronic devices.

cond-mat.mtrl-sci↗

arXiv2024

The Fusion of Large Language Models and Formal Methods for Trustworthy AI Agents: A Roadmap

Large Language Models (LLMs) have emerged as a transformative AI paradigm, profoundly influencing daily life through their exceptional language understanding and contextual generation capabilities. Despite their remarkable performance, LLMs face a critical challenge: the propensity to produce unreliable outputs due to the inherent limitations of their learning-based nature. Formal methods (FMs), on the other hand, are a well-established computation paradigm that provides mathematically rigorous techniques for modeling, specifying, and verifying the correctness of systems. FMs have been extensively applied in mission-critical software engineering, embedded systems, and cybersecurity. However, the primary challenge impeding the deployment of FMs in real-world settings lies in their steep learning curves, the absence of user-friendly interfaces, and issues with efficiency and adaptability. This position paper outlines a roadmap for advancing the next generation of trustworthy AI systems by leveraging the mutual enhancement of LLMs and FMs. First, we illustrate how FMs, including reasoning and certification techniques, can help LLMs generate more reliable and formally certified outputs. Subsequently, we highlight how the advanced learning capabilities and adaptability of LLMs can significantly enhance the usability, efficiency, and scalability of existing FM tools. Finally, we show that unifying these two computation paradigms -- integrating the flexibility and intelligence of LLMs with the rigorous reasoning abilities of FMs -- has transformative potential for the development of trustworthy AI software systems. We acknowledge that this integration has the potential to enhance both the trustworthiness and efficiency of software engineering practices while fostering the development of intelligent FM tools capable of addressing complex yet real-world challenges.

cs.AI↗

arXiv2024

ConAIR:Consistency-Augmented Iterative Interaction Framework to Enhance the Reliability of Code Generation

Code generation techniques generate code snippets automatically based on the problem requirements in natural language. Recently, large language models (LLMs) achieve the SOTA performance on code generation. However, LLMs still struggle at times to generate accurate code, which diminishes their promised efficiency as developers must spend significant effort evaluating and debugging the generated code. To improve the reliability and quality of the generated codes, researchers propose to leverage Consistency to obtain a better code based on generating and ranking multiple candidates. The existing approach is problematic as Consistency thinks a code is better when (1) the code pass more tests (inter-consistency) (2) more codes share the same behavior (intra-consistency). However, because the tests are also generated by LLMs, they could be wrong as well. As a result, majority voting based on testing results is unreliable. Relying solely on consistency is insufficient to address this issue; integrating user feedback is essential for effectively guiding consistency. We show that with minimal human effort, performance can be significantly enhanced. We propose Consistency-Augmented Iterative Interaction Framework to Enhance the Reliability of Code Generation, ConAIR, which is an approach that aims to improve the performance of a code generator through two distinctive ingredients, i.e., (1) lightweight user effort for validating the correctness of selected tests; and (2) a dynamic strategy for ranking, localizing and correcting multiple tests and codes. Overall, we propose a lightweight interaction framework that incorporates user feedback to correct identified tests and guide the iterative process. The iteration rounds are only 4 in average with the help of consistency. With only lightweight human efforts, we can achieve an improvement of 33% towards the base model.

cs.SE↗

arXiv2024

REACCEPT: Automated Co-evolution of Production and Test Code Based on Dynamic Validation and Large Language Models

Synchronizing production and test code, known as PT co-evolution, is critical for software quality in the software development lifecycle. Existing methods for automatic PT co-evolution either utilize predefined heuristic rules or rely on simple application of machine learning techniques. Due to the limitations of underlying techniques, existing methods either only partially automate PT co-evolution (e.g., only automate obsolete test code identification) or result in low accuracy. In this paper, we propose REACCEPT, a novel approach that leverages large language models and dynamic validation to fully automate PT co-evolution (i.e., capable of both identifying and updating obsolete test cases). REACCEPT relies on experience-based prompt template generation, dynamic validation, and retrieval-augmented generation techniques to accomplish automated PT co-evolution. To evaluate REACCEPT's effectiveness, we extensive experiments with a dataset of 537 Java projects and compared REACCEPT's performance with several state-of-the-art methods. Results show that REACCEPT achieved an update accuracy of 60.16% on correctly identified obsolete test code, surpassing the state-of-the-art technique CEPROT by 90%. This confirms that REACCEPT can effectively assist developers in maintaining test code, improving overall software quality and reducing maintenance effort.

cs.SE↗

arXiv2024

ALI-Agent: Assessing LLMs' Alignment with Human Values via Agent-based Evaluation

Large Language Models (LLMs) can elicit unintended and even harmful content when misaligned with human values, posing severe risks to users and society. To mitigate these risks, current evaluation benchmarks predominantly employ expert-designed contextual scenarios to assess how well LLMs align with human values. However, the labor-intensive nature of these benchmarks limits their test scope, hindering their ability to generalize to the extensive variety of open-world use cases and identify rare but crucial long-tail risks. Additionally, these static tests fail to adapt to the rapid evolution of LLMs, making it hard to evaluate timely alignment issues. To address these challenges, we propose ALI-Agent, an evaluation framework that leverages the autonomous abilities of LLM-powered agents to conduct in-depth and adaptive alignment assessments. ALI-Agent operates through two principal stages: Emulation and Refinement. During the Emulation stage, ALI-Agent automates the generation of realistic test scenarios. In the Refinement stage, it iteratively refines the scenarios to probe long-tail risks. Specifically, ALI-Agent incorporates a memory module to guide test scenario generation, a tool-using module to reduce human labor in tasks such as evaluating feedback from target LLMs, and an action module to refine tests. Extensive experiments across three aspects of human values--stereotypes, morality, and legality--demonstrate that ALI-Agent, as a general evaluation framework, effectively identifies model misalignment. Systematic analysis also validates that the generated test scenarios represent meaningful use cases, as well as integrate enhanced measures to probe long-tail risks. Our code is available at https://github.com/SophieZheng998/ALI-Agent.git

cs.AI↗

arXiv2024

Adversarial Representation Engineering: A General Model Editing Framework for Large Language Models

Since the rapid development of Large Language Models (LLMs) has achieved remarkable success, understanding and rectifying their internal complex mechanisms has become an urgent issue. Recent research has attempted to interpret their behaviors through the lens of inner representation. However, developing practical and efficient methods for applying these representations for general and flexible model editing remains challenging. In this work, we explore how to leverage insights from representation engineering to guide the editing of LLMs by deploying a representation sensor as an editing oracle. We first identify the importance of a robust and reliable sensor during editing, then propose an Adversarial Representation Engineering (ARE) framework to provide a unified and interpretable approach for conceptual model editing without compromising baseline performance. Experiments on multiple tasks demonstrate the effectiveness of ARE in various model editing scenarios. Our code and data are available at https://github.com/Zhang-Yihao/Adversarial-Representation-Engineering.

cs.LG↗

arXiv2024

Achieving Interference-Free Degrees of Freedom in Cellular Networks via RIS

It's widely perceived that Reconfigurable Intelligent Surfaces (RIS) cannot increase Degrees of Freedom (DoF) due to their relay nature. A notable exception is Jiang \& Yu's work. They demonstrate via simulation that in an ideal $K$-user interference channel, passive RIS can achieve the interference-free DoF. In this paper, we investigate the DoF gain of RIS in more realistic systems, namely cellular networks, and more challenging scenarios with direct links. We prove that RIS can boost the DoF per cell to that of the interference-free scenario even \textit{ with direct-links}. Furthermore, we \textit{theoretically} quantify the number of RIS elements required to achieve that goal, i.e. $max\left\{ {2L, (\sqrt L + c)η+L } \right\}$ (where $L=GM(GM-1)$, $c$ is a constant and $η$ denotes the ratio of channel strength) for the $G$-cells with more single-antenna users $K$ than base station antennas $M$ per cell. The main challenge lies in addressing the feasibility of a system of algebraic equations, which is difficult by itself in algebraic geometry. We tackle this problem in a probabilistic way, by exploiting the randomness of the involved coefficients and addressing the problem from the perspective of extreme value statistics and convex geometry. Moreover, numerical results confirm the tightness of our theoretical results.

cs.IT↗

arXiv2024

UniAdapt: A Universal Adapter for Knowledge Calibration

Large Language Models (LLMs) require frequent updates to correct errors and keep pace with continuously evolving knowledge in a timely and effective manner. Recent research in it model editing has highlighted the challenges in balancing generalization and locality, especially in the context of lifelong model editing. We discover that inserting knowledge directly into the model often causes conflicts and potentially disrupts other unrelated pre-trained knowledge. To address this problem, we introduce UniAdapt, a universal adapter for knowledge calibration. Inspired by the Mixture of Experts architecture and Retrieval-Augmented Generation, UniAdapt is designed with a vector-assisted router that is responsible for routing inputs to appropriate experts. The router maintains a vector store, including multiple shards, to construct routing vectors based on semantic similarity search results. UniAdapt is fully model-agnostic and designed for seamless plug-and-play integration. Experimental results show that UniAdapt outperforms existing lifelong model editors and achieves exceptional results in most metrics.

cs.LG↗

arXiv2024

Stability of the area preserving mean curvature flow in asymptotic Schwarzschild space

We first demonstrate that the area preserving mean curvature flow of hypersurfaces in space forms exists for all time and converges exponentially fast to a round sphere if the integral of the traceless second fundamental form is sufficiently small. Then we show that from sufficiently large initial coordinate sphere, the area preserving mean curvature flow exists for all time and converges exponentially fast to a constant mean curvature surface in 3-dimensional asymptotically Schwarzschild spaces. This provides a new approach to the existence of foliation established by Huisken and Yau. And also a uniqueness result follows

math.DG↗

arXiv2024

An interpretable formula for lattice thermal conductivity of crystals

Lattice thermal conductivity (kL) is a crucial physical property of crystals with applications in thermal management, such as heat dissipation, insulation, and thermoelectric energy conversion. However, accurately and rapidly determining kL poses a considerable challenge. In this study, we introduce an formula that achieves high precision (mean relative error=8.97%) and provides fast predictions, taking less than one minute, for kL across a wide range of inorganic binary and ternary materials. Our interpretable, dimensionally aligned and physical grounded formula forecasts kL values for 4,601 binary and 6,995 ternary materials in the Materials Project database. Notably, we predict undiscovered high kL values for AlBN2 (kL=101 W/ m/ K) and the undetectedlow kL Cs2Se (kL=0.98 W/ m/ K) at room temperature. This method for determining kL streamlines the traditionally time-consuming process associated with complex phonon physics. It provides insights into microscopic heat transport and facilitates the design and screening of materials with targeted and extreme kL values through the application of phonon engineering. Our findings offer opportunities for controlling and optimizing macroscopic transport properties of materials by engineering their bulk modulus, shear modulus, and Gruneisen parameter.

cond-mat.mtrl-sci↗

arXiv2024

On-the-fly Synthesis for LTL over Finite Traces: An Efficient Approach that Counts

We present an on-the-fly synthesis framework for Linear Temporal Logic over finite traces (LTLf) based on top-down deterministic automata construction. Existing approaches rely on constructing a complete Deterministic Finite Automaton (DFA) corresponding to the LTLf specification, a process with doubly exponential complexity relative to the formula size in the worst case. In this case, the synthesis procedure cannot be conducted until the entire DFA is constructed. This inefficiency is the main bottleneck of existing approaches. To address this challenge, we first present a method for converting LTLf into Transition-based DFA (TDFA) by directly leveraging LTLf semantics, incorporating intermediate results as direct components of the final automaton to enable parallelized synthesis and automata construction. We then explore the relationship between LTLf synthesis and TDFA games and subsequently develop an algorithm for performing LTLf synthesis using on-the-fly TDFA game solving. This algorithm traverses the state space in a global forward manner combined with a local backward method, along with the detection of strongly connected components. Moreover, we introduce two optimization techniques -- model-guided synthesis and state entailment -- to enhance the practical efficiency of our approach. Experimental results demonstrate that our on-the-fly approach achieves the best performance on the tested benchmarks and effectively complements existing tools and approaches.

cs.AI↗