Large Language Models (LLMs) have revolutionized natural language processing with their remarkable capabilities in text generation and reasoning. However, these models face critical challenges when deployed in real-world applications, including hallucination generation, outdated knowledge, and limited domain expertise. Retrieval And Structuring (RAS) Augmented Generation addresses these limitations by integrating dynamic information retrieval with structured knowledge representations. This survey (1) examines retrieval mechanisms including sparse, dense, and hybrid approaches for accessing external knowledge; (2) explore text structuring techniques such as taxonomy construction, hierarchical classification, and information extraction that transform unstructured text into organized representations; and (3) investigate how these structured representations integrate with LLMs through prompt-based methods, reasoning frameworks, and knowledge embedding techniques. It also identifies technical challenges in retrieval efficiency, structure quality, and knowledge integration, while highlighting research opportunities in multimodal retrieval, cross-lingual structures, and interactive systems. This comprehensive overview provides researchers and practitioners with insights into RAS methods, applications, and future directions.
论文检索
输入标题、作者或关键词，从 844 篇学术成果中精准定位
Foundation models pre-trained on massive datasets, including large language models (LLMs), vision-language models (VLMs), and large multimodal models, have demonstrated remarkable success in diverse downstream tasks. However, recent studies have shown fundamental limitations of these models: (1) limited representational capacity(2) lower adaptability, and (3) diminishing scalability. These shortcomings raise a critical question: is Euclidean geometry truly the optimal inductive bias for all foundation models, or could incorporating alternative geometric spaces enable models to better align with the intrinsic structure of real-world data and improve reasoning processes? Hyperbolic spaces, a class of non-Euclidean manifolds characterized by exponential volume growth with respect to distance, offer a mathematically grounded solution. These spaces enable low-distortion embeddings of hierarchical structures (e.g., trees, taxonomies) and power-law distributions with substantially fewer dimensions compared to Euclidean counterparts. Recent advances have leveraged these properties to enhance foundation models, including improving LLMs' complex reasoning ability, VLMs' zero-shot generalization, and cross-modal semantic alignment, while maintaining parameter efficiency. This paper provides a comprehensive review of hyperbolic neural networks and their recent development for foundation models. We further outline key challenges and research directions to advance the field.
Graph learning models have demonstrated great prowess in learning expressive representations from large-scale graph data in a wide variety of real-world scenarios. As a prevalent strategy for training powerful graph learning models, the ''pre-training, adaptation'' scheme first pre-trains graph learning models on unlabeled graph data in a self-supervised manner and then adapts them to specific downstream tasks. During the adaptation phase, graph prompting emerges as a promising approach that learns trainable prompts while keeping the pre-trained graph learning models unchanged. In this paper, we present a systematic review of recent advancements in graph prompting. First, we introduce representative graph pre-training methods that serve as the foundation step of graph prompting. Next, we review mainstream techniques in graph prompting and elaborate on how they design learnable prompts for graph prompting. Furthermore, we summarize the real-world applications of graph prompting from different domains. Finally, we discuss several open challenges in existing studies with promising future directions in this field.
Graph machine learning has led to a significant increase in the capabilities of models that learn on arbitrary graph-structured data and has been applied to molecules, social networks, recommendation systems, and transportation, among other domains. Data in multi-tabular relational databases can also be constructed as 'relational entity graphs' for Relational Deep Learning (RDL) - a new blueprint that enables end-to-end representation learning without traditional feature engineering. Compared to arbitrary graph-structured data, relational entity graphs have key properties: (i) their structure is defined by primary-foreign key relationships between entities in different tables(ii) the structural connectivity is a function of the relational schema defining a database, and (iii) the graph connectivity is temporal and heterogeneous in nature. In this paper, we provide a comprehensive review of RDL by first introducing the representation of relational databases as relational entity graphs, and then reviewing public benchmark datasets that have been used to develop and evaluate recent GNN-based RDL models. We discuss key challenges including large scale multi-table integration and the complexities of modeling temporal dynamics and heterogeneous data, while also surveying foundational neural network methods and recent architectural advances specialized for relational entity graphs. Finally, we explore opportunities to unify these distinct modeling challenges, highlighting how RDL converges multiple sub-fields in graph machine learning towards the design of foundation models that can transform the processing of relational data.
Synthetic data is often positioned as a solution to replace sensitive fixed-size datasets with a source of unlimited matching data, freed from privacy concerns. There has been much progress in synthetic data generation over the last decade, leveraging corresponding advances in machine learning and data analytics. In this survey, we cover the key developments and the main concepts in tabular synthetic data generation, including paradigms based on probabilistic graphical models and on deep learning. We provide background and motivation, before giving a technical deep-dive into the methodologies. We also address the limitations of synthetic data, by studying attacks that seek to retrieve information about the original sensitive data. Finally, we present extensions and open problems in this area.
Artificial Intelligence (AI) is rapidly reshaping the landscape of scientific discovery by enabling the development of novel models that tackle complex, data- and computation-intensive problems. Scientific challenges, in turn, provide rich, use-inspired settings that push the boundaries of AI research. This virtuous cycle is increasingly driven by cross-disciplinary collaboration, where advances in AI and domain sciences co-evolve to accelerate innovation. In this plenary panel, we will examine the opportunities and challenges in designing cutting-edge AI models for scientific discovery, and high- light the transformative potential of cross-disciplinary partnerships in shaping the future of both AI and science.
Historically, information retrieval systems have all followed the same paradigm: information seekers frame their needs in the form of a short query, the system selects a small set of relevant results from a corpus of available documents, rank-orders the results by decreasing relevance, possibly excerpts a responsive passage for each result, and returns a list of references and excerpts to the user. Retrieval systems typically did not attempt fusing information from multiple documents into an answer and displaying that answer directly. This was largely due to available technology: at the core of each retrieval system is an index that maps lexical tokens or semantic embeddings to document identifiers. Indices are designed for retrieving responsive documents; they do not support integrating these documents into a holistic answer. More recently, the coming-of-age of deep neural networks has dramatically improved the capabilities of large language models (LLMs). Trained on a large corpus of documents, these models not only memorize the vocabulary, morphology and syntax of human languages, but have shown to be able to memorize facts and relations [3]. Generative language models, when provided with a prompt, will extend the prompt with likely completions - an ability that can be used to extract answers to questions from the model. Four years ago, Metzler et al. argued that this ability of LLMs will allow us to rethink the search paradigm: to answer information needs directly rather that directing users to responsive primary sources [1]. Their vision was not without controversy; the following year Shaw and Bender argued that such a system is neither feasible nor desirable [4]. Nonetheless, the past four years have seen the emergence of such systems, with offerings from established search engines and many new entrants to the industry. This keynote (an updated version of [2]) will summarize the history of these generative information retrieval systems, and focus on the many open challenges: ensuring that answers are grounded, attributing answer passages to a primary source, providing nuanced answers to non-factoid-seeking questions, avoiding bias, and going beyond simple regurgitation of memorized facts. It will also touch on the changing nature of the content ecosystem. LLMs are starting to be used to generate web content. Should search engines treat such derived content equal to human-authored content? Is it possible to distinguish generated from original content? And how should we view hybrid authorship where humans contribute ideas and LLMs shape these ideas into prose?
AI shows promising potential to improve patient health outcomes, and the accelerating pace of technological advancement suggests health AI may be approaching a transformative threshold. Despite this promise, widespread implementation remains elusive. What barriers persist, and how might we chart a viable path forward? This talk examines two critical pathways to realizing AI's potential in healthcare: how can we enable large-scale access to clinical data and how can we evolve AI to earn the trust of medical professionals? The first challenge - lack of access to clinical data - is well-recognized but persistent. Despite recent increases in publicly available healthcare data, we still lack the volume and diversity needed to ensure accuracy across demographic groups and medical conditions. In other AI domains such as text generation, the breakthrough to reliable performance came through massive-scale training datasets. Healthcare requires a similar scale, yet patient privacy rightfully restricts data access. This presentation will explore methods for generating synthetic patient records that maintain privacy while providing the necessary training scale. The second pathway involves evolving AI to build trust within the clinical community. Medical AI must offer transparency in its decision-making processes. For instance, when answering whether ''Alice takes blood thinners'', an AI system must provide supporting evidence rather than a simple yes/no response. Two approaches will be presented that address this need: fact-verification systems for clinical claims in structured data and semantic highlighting for unstructured text. For predictive scenarios such as ''Will Alice need ventilator support in the next 48 hours?'', I will demonstrate how predictions coupled with counterfactual explanations enhance clinical trust. In addition, AI can earn provider trust by solving problems clinicians lack time to address, such as translating radiology reports into patient-friendly explanations. I will showcase methodologies that effectively bridge this communication gap. Clearing these pathways is essential to equip healthcare providers with AI tools that enable more accurate, efficient decision-making. Despite the challenges, there are compelling reasons for optimism that these barriers can be overcome, bringing us closer to truly AI-enabled healthcare.
One of the most challenging aspects of learning a foreign language is learning to converse in that language. While large language models (LLMs) have made it feasible to build generative AI chatbots that simulate conversation, creating an AI-powered language tutor that is both pedagogically effective and engaging for learners requires solving a host of additional problems. At Duolingo, we have developed an AI conversational tutor embedded in our platform, featuring a character named Lily, who helps users practice real-world conversations in their target language. In building this system, we tackled multiple challenges: adapting dialogue to each learner's proficiency level, sustaining personalized and coherent interactions across sessions, maintaining consistent character-driven personality, and designing a structure that supports both guided and learner-initiated topics. Our solution integrates a three-party conversational architecture, a persistent memory mechanism to retain prior interactions, and real-time conversation evaluation to dynamically adjust to the learner's input. This talk will highlight how generative AI, when coupled with rigorous feedback loops and thoughtful design, can significantly enhance the language learning experience.
This talk is a highly visual and accessible look at large language models, their architecture, and their training. Attendees will be presented with the intuitions for tens of LLM concepts like tokenizers, the internals of the latest Transformer neural networks, mixture-of-expert models, reward models, reasoning LLMs, model merging, and more.
Are Vision LLMs Road-Ready? A Comprehensive Benchmark for Safety-Critical Driving Video Understanding
PDF ↗Vision Large Language Models (VLLMs) have demonstrated impressive capabilities in general visual tasks such as image captioning and visual question answering. However, their effectiveness in specialized, safety-critical domains like autonomous driving remains largely unexplored. Autonomous driving systems require sophisticated scene understanding in complex environments, yet existing multimodal benchmarks primarily focus on normal driving conditions, failing to adequately assess VLLMs' performance in safety-critical scenarios. To address this, we introduce DVBench-a pioneering benchmark designed to evaluate the performance of VLLMs in understanding safety-critical driving videos. Built around a hierarchical ability taxonomy that aligns with widely adopted frameworks for describing driving scenarios used in assessing highly automated driving systems, DVBench features 10,000 multiple-choice questions with human-annotated ground-truth answers, enabling a comprehensive evaluation of VLLMs' capabilities in perception and reasoning. Experiments on 14 state-of-the-art VLLMs, ranging from 0.5B to 72B parameters, reveal significant performance gaps, with no model achieving over 40% accuracy, highlighting critical limitations in understanding complex driving scenarios. To probe adaptability, we fine-tuned selected models using domain-specific data from DVBench, achieving accuracy gains ranging from 5.24 to 10.94 percentage points, with relative improvements of up to 43.59%. This improvement underscores the necessity of targeted adaptation to bridge the gap between general-purpose vision-language models and mission-critical driving applications. DVBench establishes an essential evaluation framework and research roadmap for developing VLLMs that meet the safety and robustness requirements for real-world autonomous systems. We released the benchmark toolbox and the fine-tuned model at: https://github.com/tong-zeng/DVBench.git.
Revolutionizing Database Q&A with Large Language Models: Comprehensive Benchmark and Evaluation
PDF ↗The development of Large Language Models (LLMs) has revolutionized QA across various industries, including the database domain. However, there lacks a thorough evaluation regarding the capabilities of different LLMs in database QA. To this end, we introduce DQABench, the first comprehensive database QA benchmark for LLMs. DQABench features an innovative LLM-based method to automate the generation, cleaning, and rewriting of evaluation dataset, resulting in over 200,000 QA pairs in English and Chinese. These QA pairs cover a wide range of database-specific knowledge extracted from manuals, online communities, and DB instances, allowing for assessment of LLMs' Retrieval-Augmented Generation (RAG) and Tool Invocation Generation (TIG) capabilities in the database QA task. Furthermore, we propose a highly modular and scalable testbed DQATestbed, with basic and advanced components such as Fine-tuning, Question Classification Routing (QCR), RAG, TIG, and Prompt Template Engineering (PTE). Finally, we provide an evaluation pipeline that computes various metrics throughout a standardized evaluation process to ensure the accuracy and fairness. Our evaluation reveals the strengths and limitations of nine open-source and commercial LLMs, and the impact of various service components (e.g., fine-tuning, QCR, RAG, TIG). The proposed benchmark dataset is available at https://github.com/XMUDM/DQABench.
TH-Bench: Evaluating Evading Attacks via Humanizing AI Text on Machine-Generated Text Detectors
PDF ↗As Large Language Models (LLMs) advance, Machine-Generated Texts (MGTs) have become increasingly fluent, high-quality, and informative. Existing wide-range MGT detectors are designed to identify MGTs to prevent the spread of plagiarism and misinformation. However, adversaries attempt to humanize MGTs to evade detection (named evading attacks), which requires only minor modifications to bypass MGT detectors. Unfortunately, existing attacks generally lack a unified and comprehensive evaluation framework, as they are assessed using different experimental settings, model architectures, and datasets. To fill this gap, we introduce the Text-Humanization Benchmark (TH-Bench), the first comprehensive benchmark to evaluate evading attacks against MGT detectors. TH-Bench evaluate attacks across three key dimensions: evading effectiveness, text quality, and computational overhead. Our extensive experiments evaluate 6 state-of-the-art attacks against 13 MGT detectors across 6 datasets, spanning 19 domains and generated by 11 widely used LLMs. Our findings reveal that no single evading attack excels across all three dimensions. Through in-depth analysis, we highlight the strengths and limitations of different attacks. More importantly, we identify a trade-off among three dimensions and propose two optimization insights. Through preliminary experiments, we validate their correctness and effectiveness, offering potential directions for future research.
IVMR suite: An Industrial-scale Virtual Machine Rescheduling Dataset and Benchmark for Elastic Cloud Service
PDF ↗Virtual Machine Rescheduling (VMR) plays a crucial role in maintaining service quality and resource efficiency in elastic cloud computing. However, existing datasets and benchmarks primarily focus on VM scheduling tasks, while lacking industrial-scale datasets and standardized evaluation for the more complex and crucial rescheduling problems. To address these challenges, we present IVMR suite, the first industrial-scale suite for VMR research comprising two core components: 1) IVMR-D, an industrial-grade VMR dataset mined from a real cloud data center, integrating complete resource specifications and complex operation constraints. The dataset is systematically structured based on data size and optimization objectives. 2) IVMR-B, a benchmark for the VMR problem that establishes seamless integration of consistent evaluation and the provision of baselines spanning optimization, metaheuristic, heuristic, and machine learning-based methodologies. Our comprehensive experimental evaluation demonstrates that all tested VMR algorithms struggle to effectively balance solution quality with computational efficiency while showing limited scalability across tasks with varying complexity levels. These findings emphasize the urgency of improving VMR algorithms for industrial deployments.
Explainable AI (XAI) has gained significant attention for providing insights into the decision-making processes of deep learning models, particularly for image classification tasks through saliency-based visual explanations. Despite their success, key challenges persist due to the scarcity of annotated datasets and the absence of standardized evaluation protocols. In this paper, we introduce Saliency-Bench, a novel benchmark designed to evaluate visual explanations generated by saliency methods across multiple datasets. We curated, constructed, and annotated eight datasets, each covering diverse tasks such as scene classification, cancer diagnosis, object classification, and action classification, with corresponding ground-truth explanation annotations. The benchmark includes a standardized and unified evaluation pipeline for assessing faithfulness and alignment of the visual explanation, providing a holistic visual explanation performance assessment. We benchmark these eight datasets with widely used saliency methods on different image classifier architectures to evaluate explanation quality. Additionally, we developed an user-friendly toolkit for automating the evaluation pipeline, from data accessing, and data loading, to result evaluation. The benchmark is available at https://github.com/XAIdataset/XAIdataset.github.io.
Inspired by the success of large language models (LLMs), there is a significant research shift from traditional graph learning methods to LLM-based graph frameworks, formally known as GraphLLMs. GraphLLMs leverage the reasoning power of LLMs by integrating three key components: the textual attributes of input nodes, the structural information of node neighborhoods, and task-specific prompts that guide decision-making. Despite their promise, the robustness of GraphLLMs against adversarial perturbations remains largely unexplored-a critical concern for deploying these models in high-stakes scenarios. To bridge the gap, we introduce TrustGLM, a comprehensive study evaluating the vulnerability of GraphLLMs to adversarial attacks across three dimensions: text, graph structure, and prompt manipulations. We implement state-of-the-art attack algorithms from each perspective to rigorously assess model resilience. Through extensive experiments on six benchmark datasets from diverse domains, our findings reveal that GraphLLMs are highly susceptible to text attacks that merely replace a few semantically similar words in a node's textual attribute. We also find that standard graph structure attack methods can significantly degrade model performance, while random shuffling of the candidate label set in prompt templates leads to substantial performance drops. Beyond characterizing these vulnerabilities, we investigate defense techniques tailored to each attack vector through data-augmented training and adversarial training, which show promising potential to enhance the robustness of GraphLLMs. We hope that our open-sourced library will facilitate rapid, equitable evaluation and inspire further innovative research in this field. The benchmark code can be found in https://github.com/Palasonic5/TrustGLM.git.
As AI evolves, collaboration among heterogeneous models helps overcome data scarcity by enabling knowledge transfer across institutions and devices. Traditional Federated Learning (FL) only supports homogeneous models, limiting collaboration among clients with heterogeneous model architectures. To address this, Heterogeneous Federated Learning (HtFL) methods are developed to enable collaboration across diverse heterogeneous models while tackling the data heterogeneity issue at the same time. However, a comprehensive benchmark for standardized evaluation and analysis of the rapidly growing HtFL methods is lacking. Firstly, the highly varied datasets, model heterogeneity scenarios, and different method implementations become hurdles to making easy and fair comparisons among HtFL methods. Secondly, the effectiveness and robustness of HtFL methods are under-explored in various scenarios, such as the medical domain and sensor signal modality. To fill this gap, we introduce the first Heterogeneous Federated Learning Library (HtFLlib), an easy-to-use and extensible framework that integrates multiple datasets and model heterogeneity scenarios, offering a robust benchmark for research and practical applications. Specifically, HtFLlib integrates (1) 12 datasets spanning various domains, modalities, and data heterogeneity scenarios; (2) 40 model architectures, ranging from small to large, across three modalities; (3) a modularized and easy-to-extend HtFL codebase with implementations of 10 representative HtFL methods; and (4) systematic evaluations in terms of accuracy, convergence, computation costs, and communication costs. We emphasize the advantages and potential of state-of-the-art HtFL methods and hope that HtFLlib will catalyze advancing HtFL research and enable its broader applications. The code is released at https://github.com/TsingZ0/HtFLlib.
Large Language Models (LLMs) have revolutionized interactions between human and artificial intelligence (AI) systems, demonstrating state-of-the-art performance across various domains, including scientific discovery and hypothesis generation. However, the absence of a comprehensive and systematic evaluation framework for LLM-driven research idea generation hinders a rigorous understanding of their strengths and limitations. To address this gap, we propose IdeaBench, a benchmark system that provides a structured dataset and evaluation framework for standardizing the assessment of research idea generation by LLMs. Our dataset comprises titles and abstracts from 2,374 influential papers across eight research domains, along with their 29,408 referenced works, creating a context-rich environment that mirrors human researchers' ideation processes. By profiling LLMs as domain-specific researchers and grounding them in similar contextual constraints, we directly leverage the models' knowledge learned from the pre-training stage to generate new research ideas. To systematically evaluate LLMs' research ideation capability and approximate human assessment, we propose a reference-based metric that aligns with human judgment to quantify idea quality with the assistance of LLMs. Through this evaluation, we find that while LLMs excel at generating novel ideas, they may struggle with generating feasible ideas. IdeaBench serves as a critical resource for benchmarking and comparing LLMs, ultimately advancing research on AI's role in automating scientific discovery.
To answer complex queries on knowledge graphs, logical reasoning over incomplete knowledge needs learning-based methods because they are capable of generalizing over unobserved knowledge. Therefore, an appropriate dataset is fundamental to both obtaining and evaluating such methods under this paradigm. In this paper, we propose a comprehensive framework for data generation, model training, and method evaluation that covers the combinatorial space of Existential First-order Queries with multiple variables (EFOk). The combinatorial query space in our framework significantly extends those defined by set operations in the existing literature. Additionally, we construct a dataset, EFOk-CQA, with 741 query types for empirical evaluation, and our benchmark results provide new insights into how query hardness affects the results. Furthermore, we demonstrate that the existing dataset construction process is systematically biased and hinders the appropriate development of query-answering methods, highlighting the importance of our work. Our code and data are provided in https://github.com/HKUST-KnowComp/EFOK-CQA.
In real-world applications, graph data has garnered significant attention for its representation and analysis using Graph Neural Networks. Recent advancements have led to the development of Graph Foundation Models (GFMs), which aim to enhance cross-domain and cross-task generalization ability. Despite promising results from GFMs, a lack of standardized evaluation processes hinders comparative analysis and cross-domain applicability. To address this gap, we propose GFMBench, an open-source pipeline that standardizes the training, evaluation, and deployment of GFMs across diverse real-world graph applications. GFMBench integrates state-of-the-art GFMs and datasets, providing a modular design for comprehensive support across data preprocessing, model training, and evaluation. The pipeline includes a robust evaluation framework for benchmarking GFM generalization ability, encompassing supervised learning, cross-domain zero-shot and few-shot learning, and in-context learning. To validate the usability of GFMs, we deploy them on the Open Academic Graph, enabling applications such as topic search and author recommendation. This work provides a unified benchmark for GFMs, enabling deeper insights into their generalization ability across various graph tasks and domains. We further open-source GFMBench https://github.com/BUPT-GAMMA/ggfm and related documents https://ggfm.readthedocs.io/en/latest/.