Author name cluster

Yufeng Wang

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

15 papers

2 author rows

AAAI Conference 2026 Conference Paper

ComoRAG: A Cognitive-Inspired Memory-Organized RAG for Stateful Long Narrative Reasoning

Juyuan Wang
Rongchen Zhao
Wei Wei
Yufeng Wang
Mo Yu
Jie Zhou
Jin Xu
Liyan Xu

Narrative comprehension on long stories and novels has been a challenging domain attributed to their intricate plotlines and entangled, often evolving relations among characters and entities. Given the LLM's diminished reasoning over extended context and its high computational cost, retrieval-based approaches remain a pivotal role in practice. However, traditional RAG methods could fall short due to their stateless, single-step retrieval process, which often overlooks the dynamic nature of capturing interconnected relations within long-range context. In this work, we propose ComoRAG, holding the principle that narrative reasoning is not a one-shot process, but a dynamic, evolving interplay between new evidence acquisition and past knowledge consolidation, analogous to human cognition on reasoning with memory-related signals in the brain. Specifically, when encountering a reasoning impasse, ComoRAG undergoes iterative reasoning cycles while interacting with a dynamic memory workspace. In each cycle, it generates probing queries to devise new exploratory paths, then integrates the retrieved evidence of new aspects into a global memory pool, thereby supporting the emergence of a coherent context for the query resolution. Across four challenging long-context narrative benchmarks (200K+ tokens), ComoRAG outperforms strong RAG baselines with consistent relative gains up to 11% compared to the strongest baseline. Further analysis reveals that ComoRAG is particularly advantageous for complex queries requiring global comprehension, offering a principled, cognitively motivated paradigm for retrieval-based stateful reasoning.

PDF Details DOI

TMLR Journal 2026 Journal Article

Retrieval as a Decision: Training-Free Adaptive Gating for Efficient RAG

Yufeng Wang
Lu Wei
Haibin Ling

Retrieval-Augmented Generation (RAG) improves factuality but retrieving for every query often hurts quality while inflating tokens and latency. We propose Training-free Adaptive Retrieval Gating (\textbf{TARG}), a single-shot policy that decides when to retrieve using only a short, no-context draft from the base model. From the draft’s prefix logits, TARG computes lightweight uncertainty scores—mean token entropy, a margin signal derived from the top-1/top-2 logit gap via a monotone link, or small-$N$ variance across a handful of stochastic prefixes—and triggers retrieval only when the score exceeds a threshold. The gate is model-agnostic, adds only tens to hundreds of draft tokens, and requires no additional training or auxiliary heads. On NQ-Open, TriviaQA, and PopQA, TARG consistently pushes the accuracy–efficiency frontier: compared with Always-RAG\footnote{\textsc{Always-RAG}: retrieve for every query; \textsc{Never-RAG}: never retrieve.}, TARG matches or improves EM/F1 while reducing retrieval by 70–90\% and cutting end-to-end latency, and it remains close to Never-RAG in overhead. A central empirical finding is that under modern instruction-tuned LLMs the margin signal is a robust default (entropy compresses as backbones sharpen), with small-$N$ variance offering a conservative, budget-first alternative. We provide ablations over gate type and prefix length and use a $\Delta$-latency view to make budget trade-offs explicit.

PDF Details

IJCAI Conference 2025 Conference Paper

Efficient Dynamic Ensembling for Multiple LLM Experts

Jinwu Hu
Yufeng Wang
Shuhai Zhang
Kai Zhou
Guohao Chen
Yu Hu
Bin Xiao
Mingkui Tan

LLMs have demonstrated impressive performance across various language tasks. However, the strengths of LLMs can vary due to different architectures, model sizes, areas of training data, etc. Therefore, ensemble reasoning for the strengths of different LLM experts is critical to achieving consistent and satisfactory performance on diverse inputs across a wide range of tasks. However, existing LLM ensemble methods are either computationally intensive or incapable of leveraging complementary knowledge among LLM experts for various inputs. In this paper, we propose an efficient Dynamic Ensemble Reasoning paradigm, called DER to integrate the strengths of multiple LLM experts conditioned on dynamic inputs. Specifically, we model the LLM ensemble reasoning problem as a Markov Decision Process, wherein an agent sequentially takes inputs to request knowledge from an LLM candidate and passes the output to a subsequent LLM candidate. Moreover, we devise a reward function to train a DER-Agent to dynamically select an optimal answering route given the input questions, aiming to achieve the highest performance with as few computational resources as possible. Last, to fully transfer the expert knowledge from the prior LLMs, we develop a Knowledge Transfer Prompt that enables the subsequent LLM candidates to transfer complementary knowledge effectively. Experiments demonstrate that our method uses fewer computational resources to achieve better performance compared to state-of-the-art baselines. Code and appendix are available at https: //github. com/Fhujinwu/DER.

PDF Details DOI

EAAI Journal 2025 Journal Article

Ego-centric multiple-correlation and temporal graph neural networks based residential load forecasting

Yufeng Wang
Tianxu Han
Lingxiao Rui
Jianhua Ma
Qun Jin

Accurate Residential Load Forecasting (RLF) is pivotal for the operation and decision-making in modern power systems. Recently, as the typical implementation and application of Artificial intelligence (AI) in energy field, Graph Neural Network (GNN) based RLF has emerged as a promising paradigm, since GNN can learn from graph-structured data, and capture complex interactions among nodes in a graph. However, it is challenging to build graphs that can effectively characterize the multiple unknown dependencies among residential users. To address the above issue, this paper proposes an effective residential load forecasting framework, based on intentionally constructed multiple ego-centric networks as well as multiple correlations and temporal graph neural networks. This work's contributions are given as follows. First, from two aspects: correlation and causality, multiple personalized ego-centric networks are intentionally constructed through data-mining manner, which respectively characterize the electricity consumption similarity between households, and direct influences on ego from the ego's neighbors (so-called alters) who essentially affect the ego's RLF. Second, multiple-correlation and temporal graph neural networks are adopted to forecast the ego's load. In detail, at each timestep, the ego node's hidden feature is embedded by multiple GNNs to represent multi-correlation dependencies between the ego and its alters, then the formed feature is sent to a recurrent neural network for further learning the spatial-temporal features. Finally, thorough experiments on real datasets demonstrate that our proposal outperforms the state-of-the-art spatial-temporal GNN-based forecasting schemes. Moreover, the empirical results verify that, for the load forecasting of single ego household, data-mining based personalized graphs can indeed significantly improve the forecasting accuracy, while the formulated personalized graphs are really sparsification and locality, which reflects the intuition that there are only relatively few useful relations in graphs based RLF. The source codes are available at https: //github. com/tianxuHan/Residential-Load-Forecasting.

Details DOI

IJCAI Conference 2025 Conference Paper

Enhancing User-Oriented Proactivity in Open-Domain Dialogues with Critic Guidance

Yufeng Wang
Jinwu Hu
Ziteng Huang
Kunyang Lin
Zitian Zhang
Peihao Chen
Yu Hu
Qianyue Wang

Open-domain dialogue systems aim to generate natural and engaging conversations, providing significant practical value in real applications such as social robotics and personal assistants. The advent of large language models (LLMs) has greatly advanced this field by improving context understanding and conversational fluency. However, existing LLM-based dialogue systems often fall short in proactively understanding the user's chatting preferences and guiding conversations toward user-centered topics. This lack of user-oriented proactivity can lead users to feel unappreciated, reducing their satisfaction and willingness to continue the conversation in human-computer interactions. To address this issue, we propose a User-oriented Proactive Chatbot (UPC) to enhance the user-oriented proactivity. Specifically, we first construct a critic to evaluate this proactivity inspired by the LLM-as-a-judge strategy. Given the scarcity of high-quality training data, we then employ the critic to guide dialogues between the chatbot and user agents, generating a corpus with enhanced user-oriented proactivity. To ensure the diversity of the user backgrounds, we introduce the ISCO-800, a diverse user background dataset for constructing user agents. Moreover, considering the communication difficulty varies among users, we propose an iterative curriculum learning method that trains the chatbot from easy-to-communicate users to more challenging ones, thereby gradually enhancing its performance. Experiments demonstrate that our proposed training method is applicable to different LLMs, improving user-oriented proactivity and attractiveness in open-domain dialogues. Code and appendix are available at github. com/wang678/LLM-UPC.

PDF Details DOI

AAAI Conference 2025 Conference Paper

Graph Structure Refinement with Energy-based Contrastive Learning

Xianlin Zeng
Yufeng Wang
Yuqi Sun
Guodong Guo
Wenrui Ding
Baochang Zhang

Graph Neural Networks (GNNs) have recently gained widespread attention as a successful tool for analyzing graph-structured data. However, imperfect graph structure with noisy links lacks enough robustness and may damage graph representations, therefore limiting the GNNs' performance in practical tasks. Moreover, existing generative architectures fail to fit discriminative graph-related tasks. To tackle these issues, we introduce an unsupervised method based on a joint of generative training and discriminative training to learn graph structure and representation, aiming to improve the discriminative performance of generative models. We propose an Energy-based Contrastive Learning (ECL) guided Graph Structure Refinement (GSR) framework, denoted as ECL-GSR. To our knowledge, this is the first work to combine energy-based models with contrastive learning for GSR. Specifically, we leverage ECL to approximate the joint distribution of sample pairs, which increases the similarity between representations of positive pairs while reducing the similarity between negative ones. Refined structure is produced by augmenting and removing edges according to the similarity metrics among node representations. Extensive experiments demonstrate that ECL-GSR outperforms the state-of-the-art on eight benchmark datasets in node classification. ECL-GSR achieves faster training with fewer samples and memories against the leading baseline, highlighting its simplicity and efficiency in downstream tasks.

PDF Details DOI

NeurIPS Conference 2025 Conference Paper

Open-World Drone Active Tracking with Goal-Centered Rewards

Haowei Sun
Jinwu Hu
Zhirui Zhang
Haoyuan Tian
Xinze Xie
Yufeng Wang
Xiaohua Xie
Yun Lin

Drone Visual Active Tracking aims to autonomously follow a target object by controlling the motion system based on visual observations, providing a more practical solution for effective tracking in dynamic environments. However, accurate Drone Visual Active Tracking using reinforcement learning remains challenging due to the absence of a unified benchmark and the complexity of open-world environments with frequent interference. To address these issues, we pioneer a systematic solution. First, we propose DAT, the first open-world drone active air-to-ground tracking benchmark. It encompasses 24 city-scale scenes, featuring targets with human-like behaviors and high-fidelity dynamics simulation. DAT also provides a digital twin tool for unlimited scene generation. Additionally, we propose a novel reinforcement learning method called GC-VAT, which aims to improve the performance of drone tracking targets in complex scenarios. Specifically, we design a Goal-Centered Reward to provide precise feedback across viewpoints to the agent, enabling it to expand perception and movement range through unrestricted perspectives. Inspired by curriculum learning, we introduce a Curriculum-Based Training strategy that progressively enhances the tracking performance in complex environments. Besides, experiments on simulator and real-world images demonstrate the superior performance of GC-VAT, achieving a Tracking Success Rate of approximately 72% on the simulator. The benchmark and code are available at https: //github. com/SHWplus/DAT_Benchmark.

PDF Details

AAAI Conference 2025 Conference Paper

Open-world Radio Frequency Fingerprint Identification via Augmented Semi-supervised Learning

Zehua Han
Jing Xiao
Qirui Zhao
Zhexuan Cui
Yufeng Wang
Duona Zhang
Wenrui Ding

In complex electromagnetic environments, the identification and differentiation of diverse radio frequency (RF) emitters become particularly crucial. Existing RF fingerprinting methods demonstrate limitations when dealing with numerous unknown emitters, making it challenging for accurate classification and recognition. These limitations hinder the effective handling of specific unknown emitters.To address this issue, we introduce a novel RF fingerprinting method suitable for open-world conditions for the first time. We develop a novel RF fingerprinting model, Roinformer, to extract signal features with positional attention. We then leverage data augmentation strategies such as noise jitter and signal frame rearrangement to construct an effective pre-training model. Moreover, by incorporating instance-level similarity loss and a novel local entropy regularization approach, we significantly enhance the accuracy of known class identification and mitigate the catastrophic forgetting of known signal samples. Experimental results on three temporal signal datasets demonstrate that our method effectively recognizes both the known and unknown classes, outperforming several state-of-the-art methods by a large margin.

PDF Details DOI

IROS Conference 2024 Conference Paper

A Perceptive Pneumatic Artificial Muscle Empowered by Double Helix Fiber Reinforcement

Yufeng Wang
Houping Wu
Chenchen Li
Yulian Peng
Hongbo Wang 0002

In the last decades, soft robotics has been growing rapidly as an emerging research topic, bringing new paradigms for robotic manipulation, locomotion, and human‒machine interactions. Pneumatic artificial muscle is a powerful, lightweight, rapid response with great design flexibility, making it promising for developing biological muscle-like robotic systems. The PPAM is made of a silicone tube body with double helix coil fiber reinforcement. The double helix coil fiber restricts the radial expansion of the cylinder tube to achieve extension in actuation, and monitors the muscle length change in real time by measuring its inductance. A finite element model was built to simulate the actuation characteristics of the PPAM. A theoretical formula was derived to analyze the inductive length sensing response of the double-helix coil on the PPAM. It is verified that the PPAM can sense its length change regardless of whether it is caused by active driving or external manipulation. Rigorous testing reveals that PPAM has an ultrahigh length sensing resolution of 5. 9 μm in relaxed state, with a short response time of 50 ms. The self-length sensing of PPAM is hysteresis free, and highly repeatable, showing no degradation in 1000 operation cycles. In summary, the PPAM shows promising features for developing the next-generation perceptive and responsive soft robots, intelligent hybrid robots, or safer biomedical instruments.

Details

AAAI Conference 2023 Conference Paper

One Is All: Bridging the Gap between Neural Radiance Fields Architectures with Progressive Volume Distillation

Shuangkang Fang
Weixin Xu
Heng Wang
Yi Yang
Yufeng Wang
Shuchang Zhou

Neural Radiance Fields (NeRF) methods have proved effective as compact, high-quality and versatile representations for 3D scenes, and enable downstream tasks such as editing, retrieval, navigation, etc. Various neural architectures are vying for the core structure of NeRF, including the plain Multi-Layer Perceptron (MLP), sparse tensors, low-rank tensors, hashtables and their compositions. Each of these representations has its particular set of trade-offs. For example, the hashtable-based representations admit faster training and rendering but their lack of clear geometric meaning hampers downstream tasks like spatial-relation-aware editing. In this paper, we propose Progressive Volume Distillation (PVD), a systematic distillation method that allows any-to-any conversions between different architectures, including MLP, sparse or low-rank tensors, hashtables and their compositions. PVD consequently empowers downstream applications to optimally adapt the neural representations for the task at hand in a post hoc fashion. The conversions are fast, as distillation is progressively performed on different levels of volume representations, from shallower to deeper. We also employ special treatment of density to deal with its specific numerical instability problem. Empirical evidence is presented to validate our method on the NeRF-Synthetic, LLFF and TanksAndTemples datasets. For example, with PVD, an MLP-based NeRF model can be distilled from a hashtable-based Instant-NGP model at a 10~20X faster speed than being trained the original NeRF from scratch, while achieving a superior level of synthesis quality. Code is available at https://github.com/megvii-research/AAAI2023-PVD.

PDF Details DOI

EAAI Journal 2021 Journal Article

Attention-based learning of self-media data for marketing intention detection

Zhihao Hou
Kun Ma
Yufeng Wang
Jia Yu
Ke Ji
Zhenxiang Chen
Ajith Abraham

In the context of natural language processing, accuracy of intention detection is the basis for subsequent research on human-machine speech interaction. However, the problem of ambiguity in word vectors reduces the accuracy of intent detection. Meantime, there is a disconnection between local features and global features as well, resulting in text feature extraction that cannot fully reflect semantic information. These issues are all barriers of intention detection. Therefore, this paper proposes an attention-based convolutional neural network for self-media data learning (called A-CNN) for marketing intention. We cascade the traditional CNN with the self-attention model in the Attention networks to form a new network structure called A-CNN, and put forward a fast feature extraction method based on skip-gram-based learning called FSLText, to represent the high-dimension word vectors in the A-CNN. On the premise of maintaining the advantages of the CNN, A-CNN can not only solve the problem of local and global features disconnection caused by the CNN pooling layer, but also avoid the increase of algorithm complexity. The Self-Attention mechanism in the Attention model can effectively optimize the weight of local features of the information in global features, and retain local features that are more useful for intention detection. A fast feature extraction method which is based on Skip-gram can retain the semantic and word order information of the text. The method is beneficial to the marketing intention detection. According to the experiment, our A-CNN, compared with traditional machine learning methods, can improve 12. 32% accuracy. Contrast to the dual-channel CNN, the accuracy rate is improved by 9. 68%, and compared with the ATT-CNN, it is improved by 9. 97%. On the F1 score, the A-CNN can improve the F1 score by about 9. 37% in comparison with the traditional machine learning methods, the accuracy rate is increased by 9. 68% compared with the dual-channel CNN, and 9. 68% in contrast with ATT-CNN. It illustrates that our A-CNN can effectively address semantic and feature selection for marketing intention detection.

Details DOI

EAAI Journal 2020 Journal Article

A CLSTM-TMN for marketing intention detection

Yufeng Wang
Kun Ma
Laura Garcia-Hernandez
Jing Chen
Zhihao Hou
Ke Ji
Zhenxiang Chen
Ajith Abraham

In recent years, neural network-based models such as machine learning and deep learning have achieved excellent results in text classification. On the research of marketing intention detection, classification measures are adopted to identify news with marketing intent. However, most of current news appears in the form of dialogs. There are some challenges to find potential relevance between news sentences to determine the latent semantics. In order to address this issue, this paper has proposed a CLSTM-based topic memory network (called CLSTM-TMN for short) for marketing intention detection. A ReLU-Neuro Topic Model (RNTM) is proposed. A hidden layer is constructed to efficiently capture the subject document representation, Potential variables are applied to enhance the granularity of subject model learning. We have changed the structure of current Neural Topic Model (NTM) to add CLSTM classifier. This method is a new combination ensemble both long and short term memory (LSTM) and convolution neural network (CNN). The CLSTM structure has the ability to find relationships from a sequence of text input, and the ability to extract local and dense features through convolution operations. The effectiveness of the method for marketing intention detection is illustrated in the experiments. Our detection model has a more significant improvement in F1 (7%) than other compared models.

Details DOI

YNICL Journal 2019 Journal Article

The neural correlations of spatial attention and working memory deficits in adults with ADHD

Xiangsheng Luo
Jialiang Guo
Lu Liu
Xixi Zhao
Dongwei Li
Hui Li
Qihua Zhao
Yanfei Wang

Working memory impairment is a typical cognitive abnormality in patients with attention-deficit/hyperactivity disorder (ADHD) and is closely related to attention. Exploring the interaction between working memory and attention in patients with ADHD is of great significance for studying the pathological mechanism of this disease. In this study, electrophysiological markers of attention, posterior contralateral N2 (N2pc), and working memory, contralateral delay activity (CDA), were used to explore the relationship between these two cognitive abilities in patients with ADHD. EEG data were collected from adults with ADHD and age-, sex-, and IQ-matched normal controls while performing a classical visuospatial working memory task that consisted of low-load and high-load memory conditions. In different memory load conditions, the memory array elicited a smaller N2pc (220-260 ms) and a smaller CDA (400-800 ms) in adults with ADHD than in normal controls. Further analysis revealed that the reduced CDA amplitude could be significantly predicted by the earlier and reduced N2pc amplitude in adults with ADHD. Moreover, when the number of memory items increased, the increase in N2pc highly predicted the increases in CDA. Our findings illustrate the relationship between spatial working memory and attention ability in ADHD adults from the neurophysiological aspect that reduced working memory is closely related to insufficient attention ability and provide a potential physiological basis for the pathological mechanism of ADHD.

Details DOI

TAAS Journal 2012 Journal Article

Heterogeneity playing key role

Yufeng Wang
Akihiro Nakao
Athanasios V. Vasilakos

Heterogeneities (heterogeneous characteristics) are intrinsic in dynamic and autonomous networks, and may be caused by the following factors: finite nodes, structured network graph, mutation of node's strategy and topological view, and dynamic linking, and so on. However, few works systematically investigate the effect of the intrinsic heterogeneities on the evolutionary dynamics of incentive mechanisms in autonomous networks. In this article, we thoroughly discuss this interesting problem. Specifically, this article respectively models the pairwise interaction between peers as PD (prisoner's dilemma)-like game and multiple peers' interactions as public-goods game, proposes a general analytical framework for dynamics in evolutionary game theory (EGT)-based incentive mechanisms, and draws the following conclusions. First, for explicit incentive mechanisms, due to heterogeneity, it is impossible to get the static equilibrium of absolutely-full-cooperation (or state that provides service to the networks—so-called reciprocation), but, on the other hand, heterogeneity can facilitate reciprocation evolution, and drive the whole system into almost-full-reciprocation state, that is, most of the system time would be occupied by the full reciprocation state. Second, even without any explicit incentive mechanisms, simultaneous coevolution between dynamic linking and peers' rational strategies can not only facilitate the cooperation evolution, but drive the network structure into the desirable small-world structure. The philosophical implication of our work is that simplicity and homogeneity are too idealized for incentive mechanisms in autonomous networks—diversity and heterogeneity are intrinsic for any incentive mechanism that is compatible with the essence of our real society. Diversity is everywhere.

Details DOI

TAAS Journal 2012 Journal Article

On modeling of coevolution of strategies and structure in autonomous overlay networks

Yufeng Wang
Akihiro Nakao
Athanasios V. Vasilakos

Currently, on one hand, there exist much work about network formation and/or growth models, and on the other hand, cooperative strategy evolutions are extensively investigated in biological, economic, and social systems. Generally, overlay networks are heterogeneous, dynamic, and distributed environments managed by multiple administrative authorities, shared by users with different and competing interests, or even autonomously provided by independent and rational users. Thus, the structure of a whole overlay network and the peers' rational strategies are ever coevolving. However, there are very few approaches that theoretically investigate the coevolution between network structure and individual rational behaviors. The main motivation of our article lies in that: Unlike existing work which empirically illustrates the interaction between rational strategies and network structure (through simulations), based on EGT (Evolutionary Game Theory), we not only infer a condition that could favor the cooperative strategy over defect strategy, but also theoretically characterizes the structural properties of the formed network. Specifically, our contributions are twofold. First, we strictly derive the critical benefit-to-cost ratio ( b / c ) that would facilitate the evolution of cooperation. The critical ratio depends on the network structure (the number of peers in system and the average degree of each peer), and the evolutionary rule (the strategy and linking mutation probabilities). Then, according to the evolutionary rules, we formally derive the structural properties of the formed network in full cooperative state. Especially, the degree distribution is compatible with the power-law, and the exponent is (4-3 v )/(1-3 v ), where v is peer's linking mutation probability. Furthermore, we show that, without being harmful to cooperation evolution, a slight change of the evolutionary rule will evolve the network into a small-world structure (high global efficiency and average clustering coefficient), with the same power-law degree distribution as in the original evolution model.

Details DOI