arXiv subjects

Menghan Wang

Publications and source records attributed to Menghan Wang.

At least 19 recordsLinked to original sources

arXiv2026

Accurate and Efficient Long-Term Memory for LLM Agents

LLM agents augmented with persistent memory can recall past interactions, but existing systems suffer from two limitations: flat, unstructured storage loses relational context needed for multi-hop and temporal reasoning, and reliance on expensive LLM-based classification makes them impractical for latency-sensitive deployment. Without mechanisms to validate new information against stored knowledge, these systems silently accumulate contradictions. We present MOSAIC (Memory-Organized Structured Agent for Information Collection), a structured, conflict-aware long-term memory framework for LLM agents that is substantially more accurate and efficient. MOSAIC introduces three key capabilities: (1) entity-typed graph storage with semantic classification preserving relational structure across events, personas, and relationships, enabling multi-hop and temporal reasoning over conversation history; (2) hash-accelerated dual-path retrieval replacing LLM-based classification with locality-sensitive hashing, achieving near-instantaneous lookup with negligible accuracy loss; and (3) active conflict detection at save time that cross-references new information against existing graph neighbors, triggering updates or deletions for contradictory entries. Evaluated on LoCoMo (long-conversation QA), HaluMem, and a novel clinical-guideline error compounding test, MOSAIC achieves 89.35% accuracy on LoCoMo (+27.21 pp over the best baseline), best HaluMem-Medium extraction F1(86.77%) and HaluMem-Long extraction F1 (85.84%), best QA correctness on both Medium and Long (73.10%, 70.75%), and detects 66% of injected factual conflicts-4.7 times higher than the best baseline (14%)-while hash-accelerated retrieval keeps average search latency at 0.58 s per question.

cs.AI↗

arXiv2024

Enabling Explainable Recommendation in E-commerce with LLM-powered Product Knowledge Graph

How to leverage large language model's superior capability in e-commerce recommendation has been a hot topic. In this paper, we propose LLM-PKG, an efficient approach that distills the knowledge of LLMs into product knowledge graph (PKG) and then applies PKG to provide explainable recommendations. Specifically, we first build PKG by feeding curated prompts to LLM, and then map LLM response to real enterprise products. To mitigate the risks associated with LLM hallucination, we employ rigorous evaluation and pruning methods to ensure the reliability and availability of the KG. Through an A/B test conducted on an e-commerce website, we demonstrate the effectiveness of LLM-PKG in driving user engagements and transactions significantly.

cs.IR↗

arXiv2024

You Augment Me: Exploring ChatGPT-based Data Augmentation for Semantic Code Search

Code search plays a crucial role in software development, enabling developers to retrieve and reuse code using natural language queries. While the performance of code search models improves with an increase in high-quality data, obtaining such data can be challenging and expensive. Recently, large language models (LLMs) such as ChatGPT have made remarkable progress in both natural and programming language understanding and generation, offering user-friendly interaction via simple prompts. Inspired by these advancements, we propose a novel approach ChatDANCE, which utilizes high-quality and diverse augmented data generated by a large language model and leverages a filtering mechanism to eliminate low-quality augmentations. Specifically, we first propose a set of ChatGPT prompting rules that are specifically designed for source code and queries. Then, we leverage ChatGPT to rewrite code and queries based on the according prompts and then propose a filtering mechanism which trains a cross-encoder from the backbone model UniXcoder to filter out code and query pairs with low matching scores. Finally, we re-train the backbone model using the obtained high-quality augmented data. Experimental results show that ChatDANCE achieves state-of-the-art performance, improving the best baseline by 13.2% (R@1) and 7% (MRR). Surprisingly, we find that this augment-filter-retrain strategy enables the backbone model (UniXcoder) to self-grow. Moreover, extensive experiments show the effectiveness of each component and ChatDANCE has stable performance under different hyperparameter settings. In addition, we conduct qualitative and quantitative analyses to investigate why ChatDANCE works well and find that it learns a more uniform distribution of representations and effectively aligns the code and query spaces.

cs.SE↗

arXiv2024

Explainable Recommender with Geometric Information Bottleneck

Explainable recommender systems can explain their recommendation decisions, enhancing user trust in the systems. Most explainable recommender systems either rely on human-annotated rationales to train models for explanation generation or leverage the attention mechanism to extract important text spans from reviews as explanations. The extracted rationales are often confined to an individual review and may fail to identify the implicit features beyond the review text. To avoid the expensive human annotation process and to generate explanations beyond individual reviews, we propose to incorporate a geometric prior learnt from user-item interactions into a variational network which infers latent factors from user-item reviews. The latent factors from an individual user-item pair can be used for both recommendation and explanation generation, which naturally inherit the global characteristics encoded in the prior knowledge. Experimental results on three e-commerce datasets show that our model significantly improves the interpretability of a variational recommender using the Wasserstein distance while achieving performance comparable to existing content-based recommender systems in terms of recommendation behaviours.

cs.IR↗

arXiv2023

Modeling Orders of User Behaviors via Differentiable Sorting: A Multi-task Framework to Predicting User Post-click Conversion

User post-click conversion prediction is of high interest to researchers and developers. Recent studies employ multi-task learning to tackle the selection bias and data sparsity problem, two severe challenges in post-click behavior prediction, by incorporating click data. However, prior works mainly focused on pointwise learning and the orders of labels (i.e., click and post-click) are not well explored, which naturally poses a listwise learning problem. Inspired by recent advances on differentiable sorting, in this paper, we propose a novel multi-task framework that leverages orders of user behaviors to predict user post-click conversion in an end-to-end approach. Specifically, we define an aggregation operator to combine predicted outputs of different tasks to a unified score, then we use the computed scores to model the label relations via differentiable sorting. Extensive experiments on public and industrial datasets show the superiority of our proposed model against competitive baselines.

cs.IR↗

arXiv2023

Human Mobility Prediction with Causal and Spatial-constrained Multi-task Network

Modeling human mobility helps to understand how people are accessing resources and physically contacting with each other in cities, and thus contributes to various applications such as urban planning, epidemic control, and location-based advertisement. Next location prediction is one decisive task in individual human mobility modeling and is usually viewed as sequence modeling, solved with Markov or RNN-based methods. However, the existing models paid little attention to the logic of individual travel decisions and the reproducibility of the collective behavior of population. To this end, we propose a Causal and Spatial-constrained Long and Short-term Learner (CSLSL) for next location prediction. CSLSL utilizes a causal structure based on multi-task learning to explicitly model the "\textit{when$\rightarrow$what$\rightarrow$where}", a.k.a. "\textit{time$\rightarrow$activity$\rightarrow$location}" decision logic. We next propose a spatial-constrained loss function as an auxiliary task, to ensure the consistency between the predicted and actual spatial distribution of travelers' destinations. Moreover, CSLSL adopts modules named Long and Short-term Capturer (LSC) to learn the transition regularities across different time spans. Extensive experiments on three real-world datasets show promising performance improvements of CSLSL over baselines and confirm the effectiveness of introducing the causality and consistency constraints. The implementation is available at https://github.com/urbanmobility/CSLSL.

cs.AI↗

arXiv2023

On the Sparse DAG Structure Learning Based on Adaptive Lasso

Learning the underlying Bayesian Networks (BNs), represented by directed acyclic graphs (DAGs), of the concerned events from purely-observational data is a crucial part of evidential reasoning. This task remains challenging due to the large and discrete search space. A recent flurry of developments followed NOTEARS[1] recast this combinatorial problem into a continuous optimization problem by leveraging an algebraic equality characterization of acyclicity. However, the continuous optimization methods suffer from obtaining non-spare graphs after the numerical optimization, which leads to the inflexibility to rule out the potentially cycle-inducing edges or false discovery edges with small values. To address this issue, in this paper, we develop a completely data-driven DAG structure learning method without a predefined value to post-threshold small values. We name our method NOTEARS with adaptive Lasso (NOTEARS-AL), which is achieved by applying the adaptive penalty method to ensure the sparsity of the estimated DAG. Moreover, we show that NOTEARS-AL also inherits the oracle properties under some specific conditions. Extensive experiments on both synthetic and a real-world dataset demonstrate that our method consistently outperforms NOTEARS.

stat.ML↗

arXiv2023

Causal Inference Based Single-branch Ensemble Trees For Uplift Modeling

In this manuscript (ms), we propose causal inference based single-branch ensemble trees for uplift modeling, namely CIET. Different from standard classification methods for predictive probability modeling, CIET aims to achieve the change in the predictive probability of outcome caused by an action or a treatment. According to our CIET, two partition criteria are specifically designed to maximize the difference in outcome distribution between the treatment and control groups. Next, a novel single-branch tree is built by taking a top-down node partition approach, and the remaining samples are censored since they are not covered by the upper node partition logic. Repeating the tree-building process on the censored data, single-branch ensemble trees with a set of inference rules are thus formed. Moreover, CIET is experimentally demonstrated to outperform previous approaches for uplift modeling in terms of both area under uplift curve (AUUC) and Qini coefficient significantly. At present, CIET has already been applied to online personal loans in a national financial holdings group in China. CIET will also be of use to analysts applying machine learning techniques to causal inference in broader business domains such as web advertising, medicine and economics.

cs.LG↗

arXiv2022

Learning to Hash Naturally Sorts

Learning to hash pictures a list-wise sorting problem. Its testing metrics, e.g., mean-average precision, count on a sorted candidate list ordered by pair-wise code similarity. However, scarcely does one train a deep hashing model with the sorted results end-to-end because of the non-differentiable nature of the sorting operation. This inconsistency in the objectives of training and test may lead to sub-optimal performance since the training loss often fails to reflect the actual retrieval metric. In this paper, we tackle this problem by introducing Naturally-Sorted Hashing (NSH). We sort the Hamming distances of samples' hash codes and accordingly gather their latent representations for self-supervised training. Thanks to the recent advances in differentiable sorting approximations, the hash head receives gradients from the sorter so that the hash encoder can be optimized along with the training procedure. Additionally, we describe a novel Sorted Noise-Contrastive Estimation (SortedNCE) loss that selectively picks positive and negative samples for contrastive learning, which allows NSH to mine data semantic relations during training in an unsupervised manner. Our extensive experiments show the proposed NSH model significantly outperforms the existing unsupervised hashing methods on three benchmarked datasets.

cs.CV↗

arXiv2022

MP2: A Momentum Contrast Approach for Recommendation with Pointwise and Pairwise Learning

Binary pointwise labels (aka implicit feedback) are heavily leveraged by deep learning based recommendation algorithms nowadays. In this paper we discuss the limited expressiveness of these labels may fail to accommodate varying degrees of user preference, and thus lead to conflicts during model training, which we call annotation bias. To solve this issue, we find the soft-labeling property of pairwise labels could be utilized to alleviate the bias of pointwise labels. To this end, we propose a momentum contrast framework (MP2) that combines pointwise and pairwise learning for recommendation. MP2 has a three-tower network structure: one user network and two item networks. The two item networks are used for computing pointwise and pairwise loss respectively. To alleviate the influence of the annotation bias, we perform a momentum update to ensure a consistent item representation. Extensive experiments on real-world datasets demonstrate the superiority of our method against state-of-the-art recommendation algorithms.

cs.IR↗

arXiv2022

You Never Cluster Alone

Recent advances in self-supervised learning with instance-level contrastive objectives facilitate unsupervised clustering. However, a standalone datum is not perceiving the context of the holistic cluster, and may undergo sub-optimal assignment. In this paper, we extend the mainstream contrastive learning paradigm to a cluster-level scheme, where all the data subjected to the same cluster contribute to a unified representation that encodes the context of each data group. Contrastive learning with this representation then rewards the assignment of each datum. To implement this vision, we propose twin-contrast clustering (TCC). We define a set of categorical variables as clustering assignment confidence, which links the instance-level learning track with the cluster-level one. On one hand, with the corresponding assignment variables being the weight, a weighted aggregation along the data points implements the set representation of a cluster. We further propose heuristic cluster augmentation equivalents to enable cluster-level contrastive learning. On the other hand, we derive the evidence lower-bound of the instance-level contrastive objective with the assignments. By reparametrizing the assignment variables, TCC is trained end-to-end, requiring no alternating steps. Extensive experiments show that TCC outperforms the state-of-the-art on challenging benchmarks.

cs.CV↗

arXiv2020

M2GRL: A Multi-task Multi-view Graph Representation Learning Framework for Web-scale Recommender Systems

Combining graph representation learning with multi-view data (side information) for recommendation is a trend in industry. Most existing methods can be categorized as \emph{multi-view representation fusion}; they first build one graph and then integrate multi-view data into a single compact representation for each node in the graph. However, these methods are raising concerns in both engineering and algorithm aspects: 1) multi-view data are abundant and informative in industry and may exceed the capacity of one single vector, and 2) inductive bias may be introduced as multi-view data are often from different distributions. In this paper, we use a \emph{multi-view representation alignment} approach to address this issue. Particularly, we propose a multi-task multi-view graph representation learning framework (M2GRL) to learn node representations from multi-view graphs for web-scale recommender systems. M2GRL constructs one graph for each single-view data, learns multiple separate representations from multiple graphs, and performs alignment to model cross-view relations. M2GRL chooses a multi-task learning paradigm to learn intra-view representations and cross-view relations jointly. Besides, M2GRL applies homoscedastic uncertainty to adaptively tune the loss weights of tasks during training. We deploy M2GRL at Taobao and train it on 57 billion examples. According to offline metrics and online A/B tests, M2GRL significantly outperforms other state-of-the-art algorithms. Further exploration on diversity recommendation in Taobao shows the effectiveness of utilizing multiple representations produced by \method{}, which we argue is a promising direction for various industrial recommendation tasks of different focus.

cs.IR↗

arXiv2019

Tracing the Propagation Path: A Flow Perspective of Representation Learning on Graphs

Graph Convolutional Networks (GCNs) have gained significant developments in representation learning on graphs. However, current GCNs suffer from two common challenges: 1) GCNs are only effective with shallow structures; stacking multiple GCN layers will lead to over-smoothing. 2) GCNs do not scale well with large, dense graphs due to the recursive neighborhood expansion. We generalize the propagation strategies of current GCNs as a \emph{"Sink$\to$Source"} mode, which seems to be an underlying cause of the two challenges. To address these issues intrinsically, in this paper, we study the information propagation mechanism in a \emph{"Source$\to$Sink"} mode. We introduce a new concept "information flow path" that explicitly defines where information originates and how it diffuses. Then a novel framework, namely Flow Graph Network (FlowGN), is proposed to learn node representations. FlowGN is computationally efficient and flexible in propagation strategies. Moreover, FlowGN decouples the layer structure from the information propagation process, removing the interior constraint of applying deep structures in traditional GCNs. Further experiments on public datasets demonstrate the superiority of FlowGN against state-of-the-art GCNs.

cs.LG↗

arXiv2019

Performance of Network-Assisted Full-Duplex for Cell-Free Massive MIMO

Network assisted full-duplex (NAFD) is a spatial-division duplex technique for future wireless networks with cell-free massive multiple-input multiple-output (CF massive MIMO) network, where a large number of remote antenna units (RAUs), either using half-duplex or full-duplex, jointly support truly flexible duplex including time-division duplex, frequency-division duplex and full duplex on demand of uplink and downlink traffic by using network MIMO methods. With NAFD, bi-directional data rates of the wireless network could be increased and end-to-end delay could be reduced. In this paper, the spectral efficiency of NAFD communications in CF massive MIMO network with imperfect channel state information (CSI) is investigated under spatial correlated channels. Based on large dimensional random matrix theory, the deterministic equivalents for the uplink sum-rate with minimum-mean-square-error (MMSE) receiver as well as the downlink sum-rate with zero-forcing (ZF) and regularized zero-forcing (RZF) beamforming are derived. Numerical results show that under various environmental settings, the deterministic equivalents are accurate in both a large-scale system and system with a finite number of antennas. It is also shown that with the downlink-to-uplink interference cancellation, the uplink spectral efficiency of CF massive MIMO with NAFD could be improved. The spectral efficiencies of NAFD with different duplex configurations such as in-band full-duplex, and half-duplex are compared. With the same total numbers of transmit and receive antennas, NAFD with half-duplex RAUs offers a higher spectral efficiency. To alleviate the uplink-to-downlink interference, a novel genetic algorithm based user scheduling strategy (GAS) is proposed. Simulation results show that the achievable downlink sum-rate by using the GAS is greatly improved compared to that by using the random user scheduling.

cs.IT↗

arXiv2019

Deep Session Interest Network for Click-Through Rate Prediction

Click-Through Rate (CTR) prediction plays an important role in many industrial applications, such as online advertising and recommender systems. How to capture users' dynamic and evolving interests from their behavior sequences remains a continuous research topic in the CTR prediction. However, most existing studies overlook the intrinsic structure of the sequences: the sequences are composed of sessions, where sessions are user behaviors separated by their occurring time. We observe that user behaviors are highly homogeneous in each session, and heterogeneous cross sessions. Based on this observation, we propose a novel CTR model named Deep Session Interest Network (DSIN) that leverages users' multiple historical sessions in their behavior sequences. We first use self-attention mechanism with bias encoding to extract users' interests in each session. Then we apply Bi-LSTM to model how users' interests evolve and interact among sessions. Finally, we employ the local activation unit to adaptively learn the influences of various session interests on the target item. Experiments are conducted on both advertising and production recommender datasets and DSIN outperforms other state-of-the-art models on both datasets.

cs.IR↗

arXiv2018

Low Complexity Iterative Detection for a Large-scale Distributed MIMO Prototyping System

In this paper, we study the low-complexity iterative soft-input soft-output (SISO) detection algorithm in a large-scale distributed multiple-input multiple-output (MIMO) system. The uplink interference suppression matrix is designed to decompose the received multi-user signal into independent single-user receptions. An improved minimum-mean-square-error iterative soft decision interference cancellation (MMSE-ISDIC) based on eigenvalue decomposition (EVD-MMSE-ISDIC) is proposed to perform low-complexity detection of the decomposed signals. Furthermore, two iteration schemes are given to improve receiving performance, which are iterative detection and decoding (IDD) scheme and iterative detection (ID) scheme. While IDD utilizes the external information generated by the decoder for iterative detection, the output information of the detector is directly exploited with ID. In particular, a large-scale distributed MIMO prototyping system is introduced and a 32$ \times $32 (4 user equipments (UEs) and 4 remote antenna units (RAUs), each equipped with 8 antennas) experimental tests at 3.5 GHz was performed. The experimental results show that the proposed iterative receiver greatly outperforms the linear MMSE receiver, since it reduces the average number of error blocks of the system significantly.

eess.SP↗

arXiv2018

User-Sensitive Recommendation Ensemble with Clustered Multi-Task Learning

This paper considers recommendation algorithm ensembles in a user-sensitive manner. Recently researchers have proposed various effective recommendation algorithms, which utilized different aspects of the data and different techniques. However, the "user skewed prediction" problem may exist for almost all recommendation algorithms -- algorithms with best average predictive accuracy may cover up that the algorithms may perform poorly for some part of users, which will lead to biased services in real scenarios. In this paper, we propose a user-sensitive ensemble method named "UREC" to address this issue. We first cluster users based on the recommendation predictions, then we use multi-task learning to learn the user-sensitive ensemble function for the users. In addition, to alleviate the negative effects of new user problem to clustering users, we propose an approximate approach based on a spectral relaxation. Experiments on real-world datasets demonstrate the superiority of our methods.

cs.IR↗

arXiv2017

Collaborative Filtering with Social Exposure: A Modular Approach to Social Recommendation

This paper is concerned with how to make efficient use of social information to improve recommendations. Most existing social recommender systems assume people share similar preferences with their social friends. Which, however, may not hold true due to various motivations of making online friends and dynamics of online social networks. Inspired by recent causal process based recommendations that first model user exposures towards items and then use these exposures to guide rating prediction, we utilize social information to capture user exposures rather than user preferences. We assume that people get information of products from their online friends and they do not have to share similar preferences, which is less restrictive and seems closer to reality. Under this new assumption, in this paper, we present a novel recommendation approach (named SERec) to integrate social exposure into collaborative filtering. We propose two methods to implement SERec, namely social regularization and social boosting, each with different ways to construct social exposures. Experiments on four real-world datasets demonstrate that our methods outperform the state-of-the-art methods on top-N recommendations. Further study compares the robustness and scalability of the two proposed methods.

cs.IR↗