arXiv subjects

Liming Zhu

Publications and source records attributed to Liming Zhu.

At least 55 records · Page 3Linked to original sources

arXiv2025

International AI Safety Report 2025: First Key Update: Capabilities and Risk Implications

Since the publication of the first International AI Safety Report, AI capabilities have continued to improve across key domains. New training techniques that teach AI systems to reason step-by-step and inference-time enhancements have primarily driven these advances, rather than simply training larger models. As a result, general-purpose AI systems can solve more complex problems in a range of domains, from scientific research to software development. Their performance on benchmarks that measure performance in coding, mathematics, and answering expert-level science questions has continued to improve, though reliability challenges persist, with systems excelling on some tasks while failing completely on others. These capability improvements also have implications for multiple risks, including risks from biological weapons and cyber attacks. Finally, they pose new challenges for monitoring and controllability. This update examines how AI capabilities have improved since the first Report, then focuses on key risk areas where substantial new evidence warrants updated assessments.

cs.CY↗

arXiv2025

Ambiguity in LLMs is a concept missing problem

Ambiguity in natural language is a significant obstacle for achieving accurate text to structured data mapping through large language models (LLMs), which affects the performance of tasks such as mapping text to agentic tool calling and text-to-SQL queries. Existing methods to ambiguity handling either rely on the ReACT framework to obtain correct mappings through trial and error, or on supervised fine-tuning to bias models toward specific tasks. In this paper, we adopt a different approach that characterizes representation differences of ambiguous text in the latent space and leverages these differences to identify ambiguity before mapping them to structured data. To detect sentence-level ambiguity, we focus on the relationship between ambiguous questions and their interpretations. Unlike distances calculated by dense embeddings, we introduce a new distance measure based on a path kernel over concepts. With this measurement, we identify patterns to distinguish ambiguous from unambiguous questions. Furthermore, we propose a method for improving LLM performance on ambiguous agentic tool calling through missing concept prediction. Both achieve state-of-the-art results.

cs.CL↗

arXiv2025

HydraRAG: Structured Cross-Source Enhanced Large Language Model Reasoning

Retrieval-augmented generation (RAG) enhances large language models (LLMs) by incorporating external knowledge. Current hybrid RAG system retrieves evidence from both knowledge graphs (KGs) and text documents to support LLM reasoning. However, it faces challenges like handling multi-hop reasoning, multi-entity questions, multi-source verification, and effective graph utilization. To address these limitations, we present HydraRAG, a training-free framework that unifies graph topology, document semantics, and source reliability to support deep, faithful reasoning in LLMs. HydraRAG handles multi-hop and multi-entity problems through agent-driven exploration that combines structured and unstructured retrieval, increasing both diversity and precision of evidence. To tackle multi-source verification, HydraRAG uses a tri-factor cross-source verification (source trustworthiness assessment, cross-source corroboration, and entity-path alignment), to balance topic relevance with cross-modal agreement. By leveraging graph structure, HydraRAG fuses heterogeneous sources, guides efficient exploration, and prunes noise early. Comprehensive experiments on seven benchmark datasets show that HydraRAG achieves overall state-of-the-art results on all benchmarks with GPT-3.5-Turbo, outperforming the strong hybrid baseline ToG-2 by an average of 20.3% and up to 30.1%. Furthermore, HydraRAG enables smaller models (e.g., Llama-3.1-8B) to achieve reasoning performance comparable to that of GPT-4-Turbo. The source code is available on https://stevetantan.github.io/HydraRAG/.

cs.CL↗

arXiv2025

MMAPG: A Training-Free Framework for Multimodal Multi-hop Question Answering via Adaptive Planning Graphs

Multimodal Multi-hop question answering requires integrating information from diverse sources, such as images and texts, to derive answers. Existing methods typically rely on sequential retrieval and reasoning, where each step builds on the previous output. However, this single-path paradigm makes them vulnerable to errors due to misleading intermediate steps. Moreover, developing multimodal models can be computationally expensive, often requiring extensive training. To address these limitations, we propose a training-free framework guided by an Adaptive Planning Graph, which consists of planning, retrieval and reasoning modules. The planning module analyzes the current state of the Adaptive Planning Graph, determines the next action and where to expand the graph, which enables dynamic and flexible exploration of reasoning paths. To handle retrieval of text to unspecified target modalities, we devise modality-specific strategies that dynamically adapt to distinct data types. Our approach preserves the characteristics of multimodal information without costly task-specific training, enabling seamless integration with up-to-date models. Finally, the experiments on MultimodalQA and WebQA show that our approach matches or outperforms existing models that rely on training.

cs.AI↗

arXiv2025

Bi-level Personalization for Federated Foundation Models: A Task-vector Aggregation Approach

Federated foundation models represent a new paradigm to jointly fine-tune pre-trained foundation models across clients. It is still a challenge to fine-tune foundation models for a small group of new users or specialized scenarios, which typically involve limited data compared to the large-scale data used in pre-training. In this context, the trade-off between personalization and federation becomes more sensitive. To tackle these, we proposed a bi-level personalization framework for federated fine-tuning on foundation models. Specifically, we conduct personalized fine-tuning on the client-level using its private data, and then conduct a personalized aggregation on the server-level using similar users measured by client-specific task vectors. Given the personalization information gained from client-level fine-tuning, the server-level personalized aggregation can gain group-wise personalization information while mitigating the disturbance of irrelevant or interest-conflict clients with non-IID data. The effectiveness of the proposed algorithm has been demonstrated by extensive experimental analysis in benchmark datasets.

cs.LG↗

arXiv2025

DOCUEVAL: An LLM-based AI Engineering Tool for Building Customisable Document Evaluation Workflows

Foundation models, such as large language models (LLMs), have the potential to streamline evaluation workflows and improve their performance. However, practical adoption faces challenges, such as customisability, accuracy, and scalability. In this paper, we present DOCUEVAL, an AI engineering tool for building customisable DOCUment EVALuation workflows. DOCUEVAL supports advanced document processing and customisable workflow design which allow users to define theory-grounded reviewer roles, specify evaluation criteria, experiment with different reasoning strategies and choose the assessment style. To ensure traceability, DOCUEVAL provides comprehensive logging of every run, along with source attribution and configuration management, allowing systematic comparison of results across alternative setups. By integrating these capabilities, DOCUEVAL directly addresses core software engineering challenges, including how to determine whether evaluators are "good enough" for deployment and how to empirically compare different evaluation strategies. We demonstrate the usefulness of DOCUEVAL through a real-world academic peer review case, showing how DOCUEVAL enables both the engineering of evaluators and scalable, reliable document evaluation.

cs.IR↗

arXiv2025

SHIELDA: Structured Handling of Exceptions in LLM-Driven Agentic Workflows

Large Language Model (LLM) agentic systems are software systems powered by LLMs that autonomously reason, plan, and execute multi-step workflows to achieve human goals, rather than merely executing predefined steps. During execution, these workflows frequently encounter exceptions. Existing exception handling solutions often treat exceptions superficially, failing to trace execution-phase exceptions to their reasoning-phase root causes. Furthermore, their recovery logic is brittle, lacking structured escalation pathways when initial attempts fail. To tackle these challenges, we first present a comprehensive taxonomy of 36 exception types across 12 agent artifacts. Building on this, we propose SHIELDA (Structured Handling of Exceptions in LLM-Driven Agentic Workflows), a modular runtime exception handling framework for LLM agentic workflows. SHIELDA uses an exception classifier to select a predefined exception handling pattern from a handling pattern registry. These patterns are then executed via a structured handling executor, comprising local handling, flow control, and state recovery, to enable phase-aware recovery by linking exceptions to their root causes and facilitating composable strategies. We validate SHIELDA's effectiveness through a case study on the AutoPR agent, demonstrating effective, cross-phase recovery from a reasoning-induced exception.

cs.SE↗

arXiv2025

Context-Aware Unit Testing for Quantum Subroutines

Software testing is a critical component of the classical software development lifecycle, and this principle is expected to hold true for quantum software as it evolves toward large-scale production and adherence to industry standards. Developing and testing quantum software presents unique challenges due to the non-deterministic nature of quantum information, the high dimensionality of the underlying Hilbert space, complex hardware noise, and the inherent non-local properties of quantum systems. In this work, we model quantum subroutines as parametrized quantum channels and explore the feasibility of creating practical unit tests using probabilistic assertions, combined with either quantum tomography or statistical tests. To address the computational complexity associated with unit testing in quantum systems, we propose incorporating context-awareness into the testing process. The trade-offs between accuracy, state space coverage, and efficiency associated with the proposed theoretical framework for quantum unit testing have been demonstrated through its application to a simple three-qubit quantum subroutine that prepares a Greenberger-Horne-Zeilinger state, as well as to subroutines within a program implementing Shor's algorithm.

quant-ph↗

arXiv2025

RAGOps: Operating and Managing Retrieval-Augmented Generation Pipelines

Recent studies show that 60% of LLM-based compound systems in enterprise environments leverage some form of retrieval-augmented generation (RAG), which enhances the relevance and accuracy of LLM (or other genAI) outputs by retrieving relevant information from external data sources. LLMOps involves the practices and techniques for managing the lifecycle and operations of LLM compound systems in production environments. It supports enhancing LLM systems through continuous operations and feedback evaluation. RAGOps extends LLMOps by incorporating a strong focus on data management to address the continuous changes in external data sources. This necessitates automated methods for evaluating and testing data operations, enhancing retrieval relevance and generation quality. In this paper, we (1) characterize the generic architecture of RAG applications based on the 4+1 model view for describing software architectures, (2) outline the lifecycle of RAG systems, which integrates the management lifecycles of both the LLM and the data, (3) define the key design considerations of RAGOps across different stages of the RAG lifecycle and quality trade-off analyses, (4) highlight the overarching research challenges around RAGOps, and (5) present two use cases of RAG applications and the corresponding RAGOps considerations.

cs.SE↗

arXiv2025

LLM-based HSE Compliance Assessment: Benchmark, Performance, and Advancements

Health, Safety, and Environment (HSE) compliance assessment demands dynamic real-time decision-making under complicated regulations and complex human-machine-environment interactions. While large language models (LLMs) hold significant potential for decision intelligence and contextual dialogue, their capacity for domain-specific knowledge in HSE and structured legal reasoning remains underexplored. We introduce HSE-Bench, the first benchmark dataset designed to evaluate the HSE compliance assessment capabilities of LLM. HSE-Bench comprises over 1,000 manually curated questions drawn from regulations, court cases, safety exams, and fieldwork videos, and integrates a reasoning flow based on Issue spotting, rule Recall, rule Application, and rule Conclusion (IRAC) to assess the holistic reasoning pipeline. We conduct extensive evaluations on different prompting strategies and more than 10 LLMs, including foundation models, reasoning models and multimodal vision models. The results show that, although current LLMs achieve good performance, their capabilities largely rely on semantic matching rather than principled reasoning grounded in the underlying HSE compliance context. Moreover, their native reasoning trace lacks the systematic legal reasoning required for rigorous HSE compliance assessment. To alleviate these, we propose a new prompting technique, Reasoning of Expert (RoE), which guides LLMs to simulate the reasoning process of different experts for compliance assessment and reach a more accurate unified decision. We hope our study highlights reasoning gaps in LLMs for HSE compliance and inspires further research on related tasks.

cs.CL↗

arXiv2025

Federated Low-Rank Adaptation for Foundation Models: A Survey

Effectively leveraging private datasets remains a significant challenge in developing foundation models. Federated Learning (FL) has recently emerged as a collaborative framework that enables multiple users to fine-tune these models while mitigating data privacy risks. Meanwhile, Low-Rank Adaptation (LoRA) offers a resource-efficient alternative for fine-tuning foundation models by dramatically reducing the number of trainable parameters. This survey examines how LoRA has been integrated into federated fine-tuning for foundation models, an area we term FedLoRA, by focusing on three key challenges: distributed learning, heterogeneity, and efficiency. We further categorize existing work based on the specific methods used to address each challenge. Finally, we discuss open research questions and highlight promising directions for future investigation, outlining the next steps for advancing FedLoRA.

cs.LG↗

arXiv2025

A Qualitative Study of User Perception of M365 AI Copilot

Adopting AI copilots in professional workflows presents opportunities for enhanced productivity, efficiency, and decision making. In this paper, we present results from a six month trial of M365 Copilot conducted at our organisation in 2024. A qualitative interview study was carried out with 27 participants. The study explored user perceptions of M365 Copilot's effectiveness, productivity impact, evolving expectations, ethical concerns, and overall satisfaction. Initial enthusiasm for the tool was met with mixed post trial experiences. While some users found M365 Copilot beneficial for tasks such as email coaching, meeting summaries, and content retrieval, others reported unmet expectations in areas requiring deeper contextual understanding, reasoning, and integration with existing workflows. Ethical concerns were a recurring theme, with users highlighting issues related to data privacy, transparency, and AI bias. While M365 Copilot demonstrated value in specific operational areas, its broader impact remained constrained by usability limitations and the need for human oversight to validate AI generated outputs.

cs.CY↗

arXiv2025

CAMP in the Odyssey: Provably Robust Reinforcement Learning with Certified Radius Maximization

Deep reinforcement learning (DRL) has gained widespread adoption in control and decision-making tasks due to its strong performance in dynamic environments. However, DRL agents are vulnerable to noisy observations and adversarial attacks, and concerns about the adversarial robustness of DRL systems have emerged. Recent efforts have focused on addressing these robustness issues by establishing rigorous theoretical guarantees for the returns achieved by DRL agents in adversarial settings. Among these approaches, policy smoothing has proven to be an effective and scalable method for certifying the robustness of DRL agents. Nevertheless, existing certifiably robust DRL relies on policies trained with simple Gaussian augmentations, resulting in a suboptimal trade-off between certified robustness and certified return. To address this issue, we introduce a novel paradigm dubbed \texttt{C}ertified-r\texttt{A}dius-\texttt{M}aximizing \texttt{P}olicy (\texttt{CAMP}) training. \texttt{CAMP} is designed to enhance DRL policies, achieving better utility without compromising provable robustness. By leveraging the insight that the global certified radius can be derived from local certified radii based on training-time statistics, \texttt{CAMP} formulates a surrogate loss related to the local certified radius and optimizes the policy guided by this surrogate loss. We also introduce \textit{policy imitation} as a novel technique to stabilize \texttt{CAMP} training. Experimental results demonstrate that \texttt{CAMP} significantly improves the robustness-return trade-off across various tasks. Based on the results, \texttt{CAMP} can achieve up to twice the certified expected return compared to that of baselines. Our code is available at https://github.com/NeuralSec/camp-robust-rl.

cs.LG↗

arXiv2025

Emerging Synergies in Causality and Deep Generative Models: A Survey

In the field of artificial intelligence (AI), the quest to understand and model data-generating processes (DGPs) is of paramount importance. Deep generative models (DGMs) have proven adept in capturing complex data distributions but often fall short in generalization and interpretability. On the other hand, causality offers a structured lens to comprehend the mechanisms driving data generation and highlights the causal-effect dynamics inherent in these processes. While causality excels in interpretability and the ability to extrapolate, it grapples with intricacies of high-dimensional spaces. Recognizing the synergistic potential, we delve into the confluence of causality and DGMs. We elucidate the integration of causal principles within DGMs, investigate causal identification using DGMs, and navigate an emerging research frontier of causality in large-scale generative models, particularly generative large language models (LLMs). We offer insights into methodologies, highlight open challenges, and suggest future directions, positioning our comprehensive review as an essential guide in this swiftly emerging and evolving area.

cs.LG↗

arXiv2025

SeeAction: Towards Reverse Engineering How-What-Where of HCI Actions from Screencasts for UI Automation

UI automation is a useful technique for UI testing, bug reproduction, and robotic process automation. Recording user actions with an application assists rapid development of UI automation scripts, but existing recording techniques are intrusive, rely on OS or GUI framework accessibility support, or assume specific app implementations. Reverse engineering user actions from screencasts is non-intrusive, but a key reverse-engineering step is currently missing - recognizing human-understandable structured user actions ([command] [widget] [location]) from action screencasts. To fill the gap, we propose a deep learning-based computer vision model that can recognize 11 commands and 11 widgets, and generate location phrases from action screencasts, through joint learning and multi-task learning. We label a large dataset with 7260 video-action pairs, which record user interactions with Word, Zoom, Firefox, Photoshop, and Windows 10 Settings. Through extensive experiments, we confirm the effectiveness and generality of our model, and demonstrate the usefulness of a screencast-to-action-script tool built upon our model for bug reproduction.

cs.SE↗

arXiv2025

Privacy Bills of Materials: A Transparent Privacy Information Inventory for Collaborative Privacy Notice Generation in Mobile App Development

Privacy regulations mandate that developers must provide authentic and comprehensive privacy notices, e.g., privacy policies or labels, to inform users of their apps' privacy practices. However, due to a lack of knowledge of privacy requirements, developers often struggle to create accurate privacy notices, especially for sophisticated mobile apps with complex features and in crowded development teams. To address these challenges, we introduce Privacy Bills of Materials (PriBOM), a systematic software engineering approach that leverages different development team roles to better capture and coordinate mobile app privacy information. PriBOM facilitates transparency-centric privacy documentation and specific privacy notice creation, enabling traceability and trackability of privacy practices. We present a pre-fill of PriBOM based on static analysis and privacy notice analysis techniques. We demonstrate the perceived usefulness of PriBOM through a human evaluation with 150 diverse participants. Our findings suggest that PriBOM could serve as a significant solution for providing privacy support in DevOps for mobile apps.

cs.CR↗

arXiv2025

An Extended Pattern Collection for Blockchain-based Applications

Blockchain is an emerging technology that enables new forms of decentralized software architectures, where distributed components can reach agreements on shared system states without trusting a central integration point. Blockchain provides a shared infrastructure to execute programs, called smart contracts, and to store data. Since blockchain technologies are at an early stage, there is a lack of a systematically organized knowledge providing a holistic view on designing software systems that use blockchain. We view blockchain as a component of a bigger software system, which requires patterns for using blockchain in the design of the software architecture. In this paper, we collect a list of patterns for blockchain-based applications. The pattern collection is categorized into five categories, including interaction with external world patterns, data management patterns, security patterns, structural patterns of contracts, and user interaction patterns. Some patterns are designed considering the nature of blockchain and how blockchains can be specifically introduced within real-world applications. Others are variants of existing design patterns applied in the context of blockchain-based applications and smart contracts.

cs.SE↗

arXiv2025

Swiss Cheese Model for AI Safety: A Taxonomy and Reference Architecture for Multi-Layered Guardrails of Foundation Model Based Agents

Foundation Model (FM)-based agents are revolutionizing application development across various domains. However, their rapidly growing capabilities and autonomy have raised significant concerns about AI safety. Researchers are exploring better ways to design guardrails to ensure that the runtime behavior of FM-based agents remains within specific boundaries. Nevertheless, designing effective runtime guardrails is challenging due to the agents' autonomous and non-deterministic behavior. The involvement of multiple pipeline stages and agent artifacts, such as goals, plans, tools, at runtime further complicates these issues. Addressing these challenges at runtime requires multi-layered guardrails that operate effectively at various levels of the agent architecture. Therefore, in this paper, based on the results of a systematic literature review, we present a comprehensive taxonomy of runtime guardrails for FM-based agents to identify the key quality attributes for guardrails and design dimensions. Inspired by the Swiss Cheese Model, we also propose a reference architecture for designing multi-layered runtime guardrails for FM-based agents, which includes three dimensions: quality attributes, pipelines, and artifacts. The proposed taxonomy and reference architecture provide concrete and robust guidance for researchers and practitioners to build AI-safety-by-design from a software architecture perspective.

cs.SE↗