Accurate understanding of complex driving scenarios and the prediction of potential risks have always been critical prerequisites for autonomous vehicles to generate safe and reliable behavioral decisions. However, the uncertain states of multiple objects and their complex interactions in intricate scenes make it difficult for autonomous vehicles to extract sufficient information from representation learning to comprehend road scenarios. To address this, this paper proposes a novel method of understanding the road scene aimed at identifying potential driving risks. Firstly, we integrate the driving risk field and individual behavior features into the scene graph to enhance the dynamic characteristics of perceived targets, thereby improving the system’s sensitivity to identifying potential risks in complex scenarios. Subsequently, a causal reasoning completion module is introduced to infer and reconstruct the state of traffic entities when they are occluded or undetected, effectively enhancing the completeness of complex scene perception. Finally, we incorporate a learning framework that combines graph convolution and temporal transformers to capture node features and relationships, outputting risk prediction results. Experimental results demonstrate that the proposed method outperforms baseline approaches on both public datasets and self-constructed occlusion datasets, validating its effectiveness and superiority.

Document retrieval in real-world scenarios faces significant challenges due to diverse document formats and modalities. Traditional text-based approaches rely on tailored parsing techniques that disregard layout information and are prone to errors, while recent parsing-free visual methods often struggle to capture fine-grained textual semantics in text-rich scenarios. To address these limitations, we propose \textbf{Unveil}, a novel visual-textual embedding framework that effectively integrates textual and visual features for robust document representation. Through knowledge distillation, we transfer the semantic understanding capabilities from the visual-textual embedding model to a purely visual model, enabling efficient parsing-free retrieval while preserving semantic fidelity. Experimental results demonstrate that our visual-textual embedding method surpasses existing approaches, while knowledge distillation successfully bridges the performance gap between visual-textual and visual-only methods, improving both retrieval accuracy and efficiency.

Retrieval-Augmented Generation (RAG) systems have emerged as a pivotal methodology for enhancing Large Language Models (LLMs) through the dynamic integration of external knowledge.To further improve RAG's flexibility, Agentic RAG introduces autonomous agents into the workflow.However, Agentic RAG faces several challenges: (1) the success of each step depends on both high-quality planning and accurate search, (2) the lack of supervision for intermediate reasoning steps, and (3) the exponentially large candidate space for planning and searching.To address these challenges, we propose DecoupleSearch, a novel framework that decouples planning and search processes using dual value models, enabling independent optimization of plan reasoning and search grounding.Our approach constructs a reasoning tree, where each node represents planning and search steps.We leverage Monte Carlo Tree Search to assess the quality of each step.During inference, Hierarchical Beam Search iteratively refines planning and search candidates with dual value models.Extensive experiments across policy models of varying parameter sizes, demonstrate the effectiveness of our method.

Edge-cloud collaborative inference is a widely adopted approach in edge intelligence, especially for latency-sensitive tasks such as drone inspection, augmented reality, and disaster relief. To adapt to varying network bandwidth and limited computational capacity, model splitting methods divide Deep Neural Networks (DNNs) at specific split points into edge and cloud segments. However, current methods suffer from high switch latency and memory overhead, and are incompatible with Feature Pyramid Network (FPN)-based multi-branch models. To address the above issues, we propose a dynamic model splitting framework (DMSF) for edge-cloud collaborative inference. We design switchable compress–recover modules after each layer, enabling a single model to support all split points and significantly reduce memory overhead. The optimal split point is selected based on bandwidth and edge computational capacity, and the corresponding compress–recover path is activated for fast switching. For FPN-based multi-branch models, we remove the multi-scale branches from the backbone to the FPN and present hierarchical compensation modules in DMSF, making the model more suitable for splitting while maintaining perception accuracy. Experimental results show that DMSF reduces memory overhead by 75%, achieves millisecond-level switching, and outperforms the existing method with 32.0% lower latency and 12.2% higher perception accuracy.

Retrieval-based multimodal document QA aims to identify and integrate relevant information from visually rich documents with complex multimodal structures. While retrieval-augmented generation (RAG) has shown strong performance in text-based QA, its extensions to multimodal documents remain underexplored and face significant limitations. Specifically, current approaches rely on query-agnostic document representations that overlook salient content and use static top-k evidence selection, which fails to adapt to the uncertain distribution of relevant information. To address these limitations, we propose the Multimodal Adaptive Retrieval-Augmented (MARA) framework, which introduces query-adaptive mechanisms to both retrieval and generation. MARA consists of two components: a Query-Aligned Region Encoder that builds multi-level document representations and reweights them based on query relevance to improve retrieval precision; and a Self-Reflective Evidence Controller that monitors evidence sufficiency during generation and adaptively incorporates content from lower-ranked sources using a sliding-window strategy. Experiments on six multimodal QA benchmarks demonstrate that MARA consistently improves retrieval relevance and answer quality over existing SOTA method.

Electricity data plays a pivotal role in power management systems. Smart meters, as key tools for recording this data, often encounter anomalies due to meter malfunctions, operational errors, or unauthorized electricity usage, all of which jeopardize the stability of power grids. To this end, we propose the future-guidance anomaly detection network, called FG-Net, designed for real-time analysis of electricity time series data. FG-Net is designed to memorize historical data and assimilate future data, ensuring comprehensive learning of complete data information. Specifically, we leverage the comprehensive data insights gained from a complete information network to guide the predictions of the historical information network. Subsequently, we developed a self-matching feature guidance (SFG) strategy that harnesses the strengths of the complete information network to offset the limitations of the historical information network, thus providing effective guidance. The experimental results on two power grid time series datasets with different anomaly volatility, the Low Carbon London dataset and the Ausgrid Solar Home dataset, demonstrate the proposed anomaly detection method’s accuracy and efficiency.

Remote sensing scene graph generation aims to capture both objects and their semantic relationships, offering a comprehensive understanding of complex scenes. However, two major challenges hinder the performance of existing methods. First, remote sensing images often contain a large number of objects, many of which are unrelated. Performing global feature interactions across all objects introduces noise from irrelevant pairs, degrading feature quality. Second, relationship categories in remote sensing scenes exhibit significant intra-class variation across different contexts, and long-tailed distribution further complicates learning due to limited samples for tail classes. To address these issues, we propose the Scene-specific Multi-Prototype Network (SSMP). Our method performs contextual interactions selectively based on object and relationship categories, reducing interference from irrelevant features. Moreover, we introduce a scene-specific multi-prototype classification framework that better captures the diverse visual manifestations of each relationship class, while also improving discrimination under long-tailed distributions. Experimental results demonstrate that the proposed model achieves state-of-the-art (SOTA) performance, with a minimum improvement of 4.5% and a maximum improvement of 21.3% in mR@20 on the PredCls task over baseline models.

Drone-based remote sensing has become pivotal for high-resolution dynamic monitoring. However, the differences between day and night modes will trigger a mismatch in multi-scale object features under extreme lighting conditions. In this paper, we propose a dual-stream hierarchical interaction network for multimodal drone object detection, called DHANet, which enhances the distinguishability between multi-scale objects and background for each modality. Specifically, DHANet is designed with a Modality-Adaptive Asymmetric Attention Module (M-AAM) that enhances object-level semantic representations through global and local attention mechanisms. The M-AAM employs global context attention and local positional attention to replace conventional multi-scale context extraction, thereby effectively integrating spatial-channel information of objects. Furthermore, the network is equipped with a Multimodal Scale-Attentive Convolution (M-SC) module that dynamically generates modality-specific feature aggregation weights. This design enables global cross-modality information fusion while reducing computational complexity. Experimental results on two multimodal remote sensing benchmark datasets (DroneVehicle and VEDAI) and two natural datasets (LLVIP and FLIR) demonstrate the robustness and generalizability of DHANet. The codes will be openly and freely available at https://github.com/Victoria-xin1009/-IEEE TGRS DHANet for the sake of reproducibility.

Language-guided remote sensing image object localization uses intuitive natural language interactions to locate objects of interest within satellite or drone imagery, and has a wide range of practical applications. Early research on this task typically used discriminative models that relied on predefined task heads, limited to locating a single object and lacking flexibility. Recent years, multi-modal large language models based generative models leverage their language understanding ability to comprehend more complex language references and provide flexible outputs. However, these models tend to be too large and have limited accuracy in spotting dense, small objects in remote sensing scenarios. The reasons are: 1) Generative models treating continuous coordinate prediction as a token classification problem which fails to reflect the actual localization gap. 2) Current methods typically use CLIP pre-trained encoders that align text-image pairs at a global level, may not capture the fine-grained semantic object information necessary for accurate localization. To address these shortcomings, we introduce a lightweight generative model named LM-RSE (Localization Model with Refined Spotting Enhancement). Refined Spotting Enhancement includes the refinement of bounding box outputs and the integration of fine-grained semantic features. Specifically, we design a Bounding Box Refinement (BBR) approach that includes a special token, a Bounding Box decoder, and a custom regression loss function to refine the spotting precision and optimize the training process. Additionally, we propose a Fine-Grained Semantic Integration (FGSI) strategy, integrates a fine-grained image encoder, a Vision-Language Semantic Processing (VLSP) layer and a two-stage, full-parameter training strategy, all working together to effectively enhance the granularity of spotting. Building on this, we further refine the language-guided object localization task into two types: expression-guided and class-guided. For the former, we utilize the RSVGD dataset for evaluation and achieved state-of-the-art performance; for the latter, our evaluation results surpassed those of GeoChat. Our code and checkpoints will be released at https://github.com/Zhang-Peirong/LM-RSE.

Despite the remarkable ability of large language models (LLMs) in language comprehension and generation, they often suffer from producing factually incorrect information, also known as hallucination.A promising solution to this issue is verifiable text generation, which prompts LLMs to generate content with citations for accuracy verification.However, verifiable text generation is non-trivial due to the focus-shifting phenomenon, the intricate reasoning needed to align the claim with correct citations, and the dilemma between the precision and breadth of retrieved documents.In this paper, we present VTG, an innovative framework for Verifiable Text Generation with evolving memory and self-reflection.VTG introduces evolving long short-term memory to retain both valuable documents and recent documents.A two-tier verifier equipped with an evidence finder is proposed to rethink and reflect on the relationship between the claim and citations.Furthermore, active retrieval and diverse query generation are utilized to enhance both the precision and breadth of the retrieved documents.We conduct extensive experiments on five datasets across three knowledge-intensive tasks and the results reveal that VTG significantly outperforms baselines.

In-context learning (ICL) has been instrumental in adapting large language models (LLMs) to downstream tasks using correct input-output examples.Recent advances have attempted to improve model performance through principles derived from mistakes, yet these approaches suffer from lack of customization and inadequate error coverage.To address these limitations, we propose Retrieved In-Context Principles (RICP), a novel teacherstudent framework.In RICP, the teacher model analyzes mistakes from the student model to generate reasons and insights for preventing similar mistakes.These mistakes are clustered based on their underlying reasons for developing task-level principles, enhancing the error coverage of principles.During inference, the most relevant mistakes for each question are retrieved to create question-level principles, improving the customization of the provided guidance.RICP is orthogonal to existing prompting methods and does not require intervention from the teacher model during inference.Experimental results across seven reasoning benchmarks reveal that RICP effectively enhances performance when applied to various prompting strategies.

Embedding-based retrieval methods construct vector indices to search for document representations that are most similar to the query representations. They are widely used in document retrieval due to low latency and decent recall performance. Recent research indicates that deep retrieval solutions offer better model quality, but are hindered by unacceptable serving latency and the inability to support document updates. In this paper, we aim to enhance the vector index with end-to-end deep generative models, leveraging the differentiable advantages of deep retrieval models while maintaining desirable serving efficiency. We propose Model-enhanced Vector Index (MEVI), a differentiable model-enhanced index empowered by a twin-tower representation model. MEVI leverages a Residual Quantization (RQ) codebook to bridge the sequence-to-sequence deep retrieval and embedding-based models. To substantially reduce the inference time, instead of decoding the unique document ids in long sequential steps, we first generate some semantic virtual cluster ids of candidate documents in a small number of steps, and then leverage the well-adapted embedding vectors to further perform a fine-grained search for the relevant documents in the candidate virtual clusters. We empirically show that our model achieves better performance on the commonly used academic benchmarks MSMARCO Passage and Natural Questions, with comparable serving latency to dense retrieval solutions.

Current state-of-the-art document retrieval solutions mainly follow an index-retrieve paradigm, where the index is hard to be directly optimized for the final retrieval target. In this paper, we aim to show that an end-to-end deep neural network unifying training and indexing stages can significantly improve the recall performance of traditional methods. To this end, we propose Neural Corpus Indexer (NCI), a sequence-to-sequence network that generates relevant document identifiers directly for a designated query. To optimize the recall performance of NCI, we invent a prefix-aware weight-adaptive decoder architecture, and leverage tailored techniques including query generation, semantic document identifiers, and consistency-based regularization. Empirical studies demonstrated the superiority of NCI on two commonly used academic benchmarks, achieving +21.4% and +16.8% relative enhancement for Recall@1 on NQ320k dataset and R-Precision on TriviaQA dataset, respectively, compared to the best baseline method.