The development of large language models (LLMs) has achieved superior performance in a range of downstream tasks, including LLM-based retrieval-augmented generation (RAG). The quality of generated content heavily relies on the usefulness of the retrieved information and the capacity of LLMs' internal information processing mechanism to incorporate it in answer generation. It is generally assumed that the retrieved information is relevant to the question. However, the retrieved information may have a variable degree of relevance and usefulness, depending on the question and the document collection. It is important to take into account the relevance of the retrieved information in answer generation. In this paper, we propose OpenDecoder, a new approach that leverages explicit evaluation of the retrieved information as quality indicator features for generation. We aim to build a RAG model that is more robust to varying levels of noisy context. Three types of explicit evaluation information are considered: relevance score, ranking score, and QPP (query performance prediction) score. The experimental results on five benchmark datasets demonstrate the effectiveness and better robustness of OpenDecoder by outperforming various baseline methods. Importantly, this paradigm is flexible to be integrated with the post-training of LLMs for any purposes and incorporated with any type of external indicators.
论文检索
输入标题、作者或关键词，从 8,216 篇学术成果中精准定位
Large Language Models (LLMs) have demonstrated strong capabilities in web search and reasoning. However, their dependence on static training corpora makes them prone to factual errors and knowledge gaps. Retrieval-Augmented Generation (RAG) addresses this limitation by incorporating external knowledge sources, especially structured Knowledge Graphs (KGs), which provide explicit semantics and efficient retrieval. Existing KG-based RAG approaches, however, generally assume that anchor entities are accessible to initiate graph traversal, which limits their robustness in open-world settings where accurate linking between the user query and the KG entity is unreliable. To overcome this limitation, we propose AnchorRAG, a novel multi-agent collaboration framework for open-world RAG without the predefined anchor entities. Specifically, a predictor agent dynamically identifies candidate anchor entities by aligning user query terms with KG nodes and initializes independent retriever agents to conduct parallel multi-hop explorations from each candidate. Then a supervisor agent formulates the iterative retrieval strategy for these retriever agents and synthesizes the resulting knowledge paths to generate the final answer. This multi-agent collaboration framework improves retrieval robustness and mitigates the impact of ambiguous or erroneous anchors. Extensive experiments on four public benchmarks demonstrate that AnchorRAG significantly outperforms existing baselines and establishes new state-of-the-art results on the real-world reasoning tasks.
Composed Image Retrieval (CIR) aims to retrieve target images based on a reference image and modified texts. However, existing methods often struggle to extract the correct semantic cues from the reference image that best reflect the user's intent under textual modification prompts, resulting in interference from irrelevant visual noise. In this paper, we propose a novel Multi-level Vision Selection by Multi-modal Chain-of-Thought Reasoning (MCoT-MVS) for CIR, integrating attention-aware multi-level vision features guided by reasoning cues from a multi-modal large language model (MLLM). Specifically, we leverage an MLLM to perform chain-of-thought reasoning on the multimodal composed input, generating the retained, removed, and target-inferred texts. These textual cues subsequently guide two reference visual attention selection modules to selectively extract discriminative patch-level and instance-level semantics from the reference image. Finally, to effectively fuse these multi-granular visual cues with the modified text and the imagined target description, we design a weighted hierarchical combination module to align the composed query with target images in a unified embedding space. Extensive experiments on two CIR benchmarks, namely CIRR and FashionIQ, demonstrate that our approach consistently outperforms existing methods and achieves new state-of-the-art performance. Code and trained models are publicly released at https://github.com/JJJJerry/WWW2026-MCoT-MVS.
Cross-modal retrieval (CMR) has advanced considerably by mapping image and text features into a shared embedding space; however, these approaches still face two persistent challenges: (1) semantic sparsity, where discriminative cues are confined to localized regions, making it difficult to identify implicit visual evidence; and (2) ranking uncertainty under semantic ambiguity, where models struggle to maintain the correct retrieval order when candidates share similar contexts. To address these issues, we propose the Dual-Branch Multi-Granularity Network (DBMG) with Structured Contrastive Ranking, which enriches visual semantics by leveraging a multimodal large language model to generate auxiliary descriptions, aligns sparse cues through a dual-branch architecture capturing both global and local interactions, and enforces ranking consistency via a three-stage contrastive objective that progressively optimizes category clustering, instance alignment, and margin-based ranking. Extensive experiments on four standard CMR benchmarks demonstrate that DBMG outperforms 12 strong baselines, achieving an average 15.91% improvement in mAP, establishing a new state-of-the-art. The code is available at https://github.com/DMiC-Lab-HFUT/DBMG.
Long videos contain a vast amount of information, making video-text retrieval an essential and challenging task in multimodal learning and web-scale search. On today's Web, where users increasingly expect to locate not only relevant pages but also specific long videos or fine-grained clips, existing benchmarks fall short due to limited video duration, low-quality captions, and coarse annotation granularity. To address these limitations, we introduce LoVR, a benchmark specifically designed for long video-text retrieval. LoVR contains 467 long videos and over 40,804 fine-grained clips with high-quality captions. To overcome the issue of poor machine-generated annotations, we propose an efficient caption generation framework that integrates VLM automatic generation, caption quality scoring, and dynamic refinement. This pipeline improves annotation accuracy while maintaining scalability. Furthermore, we introduce a semantic fusion method to generate coherent full-video captions without losing important contextual information. Our benchmark introduces longer videos, more detailed captions, and a larger-scale dataset, presenting new challenges for video understanding and retrieval. Extensive experiments on various advanced models demonstrate that LoVR is a challenging benchmark, revealing the limitations of current approaches and providing valuable insights for future research. We release the code link at https://github.com/TechNomad-ds/LoVR-benchmark/.
The dominant paradigm for interacting with the Web has long relied on menu-driven navigation and complex graphical structures that can impose a high cognitive load on users. Recent advances in Large Language Models (LLMs) enable flexible NL understanding, yet purely conversational systems lack the stability and visual affordances required for complex tasks. This paper introduces Mirage, a framework that bridges these paradigms by making graphical user interface (GUI) components interpretable and orchestrable by LLMs, enabling on-the-fly construction of simplified UIs tailored to user requests. Developers declare application functions as reusable components with semantic descriptions and typed schemas; Mirage maps user goals, expressed in natural language, to the most relevant components and renders them as interactive interfaces. We implemented Mirage as a Next.js framework and validated it in a user study comparing the new paradigm to the navigation of a traditional website. Results show that Mirage significantly reduces task completion time, lowers cognitive workload, and achieves higher usability without compromising accuracy. Our approach also highlights new design directions toward supporting on-the-fly building of goal-driven interfaces.
Data-driven spectrum sensing is a key technology for addressing complex challenges in Cognitive Radio Networks (CRNs). Traditional methods are typically designed for simple single-band scenarios and perform poorly in practical wideband applications. In real-world systems, a single Secondary User (SU) is often restricted by energy, time, and hardware capabilities during real-time sensing. Consequently, only local and fragmented frequency information can be obtained. This partial sensing leads to a severe lack of training data. Additionally, the lack of historical records for emerging frequency bands, combined with data incompleteness due to resource constraints, creates training bottlenecks for data-driven models and limits the reliability of sensing. To address these challenges, this paper proposes a novel framework based on Collaborative Subgraph Learning and Hyperbolic Graph Neural Networks (GNNs). This approach enables Secondary Users to perform collaborative sensing through distributed subgraph learning. By utilizing GNNs to extract features and model multi-band correlations, a new distributed GNNs architecture is designed to efficiently detect wideband spectrum occupancy, even with partial observations. Within this framework, all frequency bands in the wideband spectrum pool are treated as a unified graph, while the bands observed by each SU form a subgraph. Subsequently, the complete spectrum graph is constructed through the joint training and aggregation of these subgraphs. By integrating hyperbolic geometry into GNNs, this method better captures the hierarchical structure of spectrum patterns, providing a more accurate and efficient sensing model. Experimental results demonstrate that, compared to the second-best HCNNs model, the proposed framework improves sensing accuracy by 3.8% on average across various test environments, while reducing key resource consumption by 18.4% on average.
Graph generation is a fundamental problem in graph learning with broad applications across Web-scale systems, knowledge graphs, and scientific domains such as drug and material discovery. Recent approaches leverage diffusion models for step-by-step generation, yet unconditional diffusion offers little control over desired properties, often leading to unstable quality and difficulty in incorporating new objectives. Inference-time guidance methods mitigate these issues by adjusting the sampling process without retraining, but they remain inherently local, heuristic, and limited in controllability. To overcome these limitations, we propose TreeDiff, a Monte Carlo Tree Search (MCTS) guided dual-space diffusion framework for controllable graph generation. TreeDiff is a plug-and-play inference-time method that expands the search space while keeping computation tractable. Specifically, TreeDiff introduces three key designs to make it practical and scalable: (1) a macro-step expansion strategy that groups multiple denoising updates into a single transition, reducing tree depth and enabling long-horizon exploration; (2) a dual-space denoising mechanism that couples efficient latent-space denoising with lightweight discrete correction in graph space, ensuring both scalability and structural fidelity; and (3) a dual-space verifier that predicts long-term rewards from partially denoised graphs, enabling early value estimation and removing the need for full rollouts. Extensive experiments on 2D and 3D molecular generation benchmarks, under both unconditional and conditional settings, demonstrate that TreeDiff achieves state-of-the-art performance. Notably, TreeDiff exhibits favorable inference-time scaling: it continues to improve with additional computation, while existing inference-time methods plateau early under limited resources.
Cross-domain recommendation (CDR) has emerged as a promising solution by effectively alleviating data sparsity by leveraging information from auxiliary domains. However, a major challenge in CDR is its dependence on predefined alignment rules (e.g., structural or distribution matching) to achieve cross-domain knowledge transfer, which impose fixed transfer patterns and lack the flexibly need for diverse cross-domain scenarios. Furthermore, most existing approaches still rely on coarse-grained representations. Knowledge transfer built upon imprecise representations can, even with improved alignment rules, instead lead to negative transfer in the target domain. To address these challenges and optimize recommendation efficacy, a new framework named meta-learning driven few-shot knowledge transfer with dual-stage adaptive data replay for cross-domain recommendation (MFACDR) is proposed. Specifically, a new meta-learning driven few-shot knowledge transfer method is proposed. This method leverages overlapping parts as anchors to guide the non-overlapping parts in autonomously exploring alignment rules through meta-learning, thus enabling few-shot knowledge transfer and flexible handling of different cross-domain scenarios. In addition, a dual-stage adaptive data replay mechanism is proposed, which enables fine-grained cross-domain adaptability and helps to mitigate negative transfer. Extensive experiments on three real-world datasets consistently demonstrate the superior effectiveness and robustness of the proposed MFACDR.
Dynamic graph representation learning has become essential for analyzing evolving networks in domains such as social network analysis, recommendation systems, and traffic analysis. However, existing continuous-time methods face three key challenges: (1) some methods depend solely on node-specific memory without effectively incorporating information from neighboring nodes, resulting in embedding staleness; (2) most fail to explicitly capture correlations between node neighborhoods, limiting contextual awareness; and (3) many fail to fully capture the structural dynamics of evolving graphs, especially in absence of rich link attributes. To address these limitations, we introduce TAWRMAC—a novel framework that integrates Temporal Anonymous Walks with Restart, Memory Augmentation, and Neighbor Co-occurrence embedding. TAWRMAC enhances embedding stability through a memory-augmented GNN with fixed-time encoding and improves contextual representation by explicitly capturing neighbor correlations. Additionally, its Temporal Anonymous Walks with Restart mechanism distinguishes between nodes exhibiting repetitive interactions and those forming new connections beyond their immediate neighborhood. This approach captures structural dynamics better and supports strong inductive learning. Extensive experiments on multiple benchmark datasets demonstrate that TAWRMAC consistently outperforms state-of-the-art methods in dynamic link prediction and node classification under both transductive and inductive settings across three different negative sampling strategies. By providing stable, generalizable, and context-aware embeddings, TAWRMAC advances the state of the art in continuous-time dynamic graph learning.
State-of-the-art algorithms for randomizing (or ''null-modeling'') graphs, hypergraphs, binary matrices, and market basket datasets are based on the Curveball trade, a procedure that takes two n-dimensional binary vectors (e.g., 1-hop neighborhood vectors, hyperedges, columns, transactions), and transforms them into two new vectors with the same L1 norm, bitwise-XOR, and bitwise-AND as the original ones. We present HomeRun, an algorithm to perform Curveball trades in an almost-streaming fashion. Previous approaches make multiple passes over the input vectors, and perform other operations with O(n) computational cost\@. HomeRun instead performs a single linear scan over the data, and costs less than two passes. We achieve this speedup by framing a Curveball trade as a sampling procedure, and by using reservoir sampling and strict upper bounds to avoid having to compute the population and sample sizes. Our experimental evaluation on real and artificial datasets shows that HomeRun achieves a 2x speedup over existing algorithms.
The diffusion model with multiple conditions has received widespread attention in the field of drug design due to its high-quality generation ability. However, the paradigm of directly generating new molecules from conditions used in existing work has not accurately fitted the joint distribution of multiple conditions during the generation process. To address this issue, we propose Graph Diffusion Evolution Model(GDEM) for multi conditional molecule generation. GDEM decomposes the process of molecular generation into a chain-like Markov evolution process, continuously adjusting the molecular structure and gradually approaching the true multi-conditional joint distribution. Meanwhile, in order to effectively train this chain evolution generative model, we also propose a two-stage training approximation method to complete the training of intermediate steps. We validated the effectiveness of GDEM on multiple polymer datasets and small molecule datasets, and the results showed that GDEM has advantages in molecular properties and condition control compared to traditional methods.
Web3 technologies and decentralised finance (DeFi) have not only revolutionised finance, but they have also given rise to completely different concepts in terms of how we handle our wealth and how we utilise the Internet, which is still an unreliable medium for the exchange of information. To overcome trust issues, cryptographic tools were introduced and these days we even put our digital assets on the blockchain. These new technologies enabled us to make most financial solutions decentralised. A widely popular financial product that is implemented on blockchains is a constant product market maker (CPMM). A CPMM can be used to swap digital assets between mutually untrusting parties by relying on prices calculated from its bonding curve. A blockchain-based automated market maker (AMM) quotes prices that usually lag behind external market prices, and this price difference can be exploited by automated bots, aka. arbitrageurs. The question we investigate in this work focuses on this issue. More specifically, we experiment with different approaches from financial mathematics to account for stochastic price changes, perform extensive measurements to quantify the characteristics of the DeFi market, focusing on mainnet Ethereum and develop stochastic grid optimisers to fit model parameters to actual AMM market data that we extract from the logs of the Ethereum virtual machine. We report on model performance that is more aligned with reality compared to previous works that have relied on naive two-parameter Black-Scholes (BS) models. Our results reveal that hidden Markov models can give better estimates for both the number of arbitrage trades and the arbitrage profits when combined with a jump diffusion model and a two-parameter BS model.
We often treat data as an infinitely reusable resource: a single dataset can support many analyses, train multiple models, and be shared widely without apparent cost. This talk argues that in important ways, data is not endlessly reusable. Instead, in certain contexts, data behaves like a consumable resource that degrades with use. The clearest example arises in the presence of privacy concerns. Fundamental results show that any informative public analysis of personal data inevitably leaks some information about the underlying individuals, and that these privacy losses accumulate across repeated uses of the same or overlapping datasets [1]. If some level of privacy is to be preserved, this imposes intrinsic limits on how many times data can be used. In joint work under submission, we connect this perspective to the mosaic effect from legal scholarship, arguing that privacy risks arise not only from combining data pieces, but also from combining seemingly innocuous data uses [6]. This view suggests regulatory and technical approaches that treat data use itself as a rival good [4]. Data can also degrade with use even when privacy is not at stake. A line of work on adaptive data analysis shows that repeatedly querying the same dataset can lead to overfitting: results that appear valid on the dataset but fail to generalize to the underlying distribution, even when the dataset is very large [2, 3, 5]. In both privacy and generalization, each interaction with a dataset consumes part of a limited resource, constraining future computations. Recognizing data degradation opens a range of research directions, including systems for tracking and budgeting data use, algorithmic techniques to mitigate degradation, the role of synthetic data and data curators, and new models of non-worst-case adaptive computation. Together, these directions work towards a data ecosystem that explicitly accounts for data degradation.
The performance of modern AI systems is fundamentally constrained by the quality of their underlying GPU kernels, which translate high-level algorithmic semantics into low-level hardware operations. Achieving near-optimal kernels requires expert-level understanding of hardware architectures and programming models, making kernel engineering a critical but notoriously time-consuming and non-scalable process. Recent advances in large language models and LLM-based agents have opened new possibilities for automating kernel generation and optimization. LLMs are well-suited to compress expert-level kernel knowledge that is difficult to formalize, while agentic systems further enable scalable optimization by casting kernel development as an iterative, feedback-driven loop. Rapid progress has been made in this area. However, the field remains fragmented and lacks a systematic perspective for LLM-driven kernel generation. This survey addresses this gap by providing a structured overview of existing approaches, spanning LLM-based approaches and agentic optimization workflows, and systematically organizing the datasets and benchmarks that underpin learning and evaluation in this domain. Moreover, key open challenges and future research directions are further outlined, aiming to establish a comprehensive reference for the next generation of automated kernel optimization. To keep track of this field, we maintain an open-source GitHub repository at https://github.com/flagos-ai/awesome-LLM-driven-kernel-generation.
Artificial intelligence (AI) applications are prone to reproducing existing patterns of marginalization, bias, and discrimination. The adoption of automated decision-making tools, particularly those based on large language models (LLMs), necessitates the detection and analysis of their discriminatory behavior to ensure compliance with anti-discrimination regulations. This work advances metamorphic testing for bias detection, aiming to (1) uncover discriminatory behavior in AI tools and (2) standardize the metamorphic test creation approach, rooting it in the legal and academic research on discrimination. To fill in these gaps, we introduce a metamorphic test creation workflow and evaluate it on the LLM-based CV screening use case. Our method yields a metamorphic relation (MR) catalogue of 8 atomic and 6 composite MRs, legally justified by Article 21 (1) of the Charter of Fundamental Rights of the European Union and sociologically motivated by the intersectionality theory. The resulting metamorphic test suite uncovers discriminatory failures for all 6 evaluated LLMs, totaling 1,272 defects across 7,800 test cases. Both the new atomic MRs and the composite MRs detect additional discriminatory defects for all LLMs. Composite MRs consistently uncovered more defects than their atomic counterparts, with failure rates increasing by on average 9% for Llama-3.1 8B and 25% for Llama-3.2 3B.
Forecasting Origin-Destination (OD) demand for new metro lines is critical for sustainable infrastructure planning but faces spatiotemporal out-of-distribution challenges. Existing models often struggle to capture heterogeneous interaction patterns in changing topologies and overlook inherent uncertainty and over-dispersion issues. To bridge these gaps, we propose PROB-EMOE, a planning-oriented probabilistic framework tailored for network expansion. To ensure robust generalization, we design a Mixture-of-Experts (MoE) predictor that integrates diverse expert views to capture heterogeneous demand patterns across changing topologies. To quantify extrapolation uncertainty, our framework functions as a unified probabilistic system by synergizing Deep Ensembles with a probabilistic output, effectively quantifying both data and model uncertainty. Through a systematic investigation of likelihood families, we empirically demonstrate that the Negative Binomial distribution offers the optimal fit in this context. Extensive experiments on a multi-year Shenzhen metro dataset demonstrate that our approach achieves state-of-the-art predictive performance and provides the sharpest calibrated uncertainty intervals. The framework has been deployed in a metropolitan smart-data platform to support risk-aware investment decisions.
Along with the rapid development of new urban mobility options like ride-sharing over the past decade, on-demand micro-transit services stand out as a middle ground, bridging the gap between fixed-line mass transit and single-request ride-hailing, balancing ridership maximization and travel time minimization. Micro-transit adoption can have significant social impact. It improves urban sustainability, through lower energy consumption and reduced emissions, while enhancing equitable mobility access for disadvantaged communities, thanks to its lower vehicle miles per passenger, flexible schedules, and affordable pricing. However, effective operation of micro-transit services requires planning geo-fenced zones in advance, which involves solving a challenging combinatorial optimization problem. Existing approaches enumerate candidate zones first and selects a fixed number of optimal zones in the second step. In this paper, we generalize the Micro-Transit Zoning Problem (MZP) to allow a global budget rather than imposing a size limit for candidate zones. We also design a Column Generation (CG) framework to solve the problem exactly and several pricing heuristics to accelerate computation. Extensive numerical experiments across major U.S. cities demonstrate that our approach produces higher-quality solutions more efficiently and scales better in the generalized setting.
Understanding and monitoring human behavior in metro stations play an important role in supporting suicide prevention efforts, where early identification of high-risk situations can enable timely intervention. This requires assessing suicide risk from a surveillance video by jointly reasoning about the behavior of each passenger, his/her spatial context, and temporal dynamics. However, this assessment using videos captured by surveillance cameras is challenging, as it demands accurate perception of human motion, understanding of platform geometry, and aggregation of heterogeneous behavioral cues over time. In this work, we formalize the task of Suicide Risk Assessment (SRA) in metro stations and introduce the first interpretable framework that addresses this challenge. Unlike approaches that focus on isolated subtasks or attempt to infer intent directly, our formulation assesses suicide risk from accumulated evidence by incorporating person tracking, activity recognition, semantic segmentation of the platform, and trajectory-driven risk heatmap modeling. By formalizing SRA as a distinct task and benchmarking a complete operational pipeline achieving 83.2% ROC-AUC on real surveillance data, this work highlights the complexity of suicide risk assessment and opens new directions for research on interpretable AI systems for social good.
Temporal constraints are an intrinsic component of most clinical guidelines. Several approaches to computerized clinical guidelines (CIGs) offer temporal reasoning facilities to support the execution of a CIG for a specific patient, mostly based on bounds on differences and on the Simple Temporal Problem framework. However, in scheduling activities (e.g., within a hospital), it is necessary to consider multiple executions of CIGs for different patients, which can also be (partly) related to each other. In this work we extend current temporal reasoning techniques to apply to such a context. We propose a new temporal constraint model, prove its properties, and exploit them to provide efficient management of patients' temporal constraints, and to support efficient query answering. We also propose an experimental evaluation, demonstrating the step forward with respect to current approaches. Notably, our approach is general, and can apply to all temporal reasoning problems having the same topology of the "multiple CIG execution" problem.