Chuhan Wang, Xintong Li, Jennifer Yuntong Zhang, Junda Wu, Chengkai Huang, Lina Yao, Julian McAuley, Jingbo Shang. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Large language models (LLMs) have recently been adopted for recommendation by framing user preference modeling as a language generation problem. However, existing latent reasoning approaches typically represent user intent with a single latent vector, which struggles to capture the inherently multi-faceted nature of user preferences. We propose Factorized Latent Reasoning (FLR), a novel framework for LLM-based sequential recommendation that decomposes latent reasoning into multiple disentangled preference factors. FLR introduces a lightweight multi-factor attention module that iteratively refines a latent thought representation, where each factor attends to distinct aspects of the user's interaction history. To encourage diversity and specialization, we design orthogonality, attention diversity, and sparsity regularization objectives, and dynamically aggregate factor contributions for the final prediction. We further integrate FLR with an efficient reinforcement learning strategy based on group-relative policy optimization, enabling stable alignment directly in the latent reasoning space. Experiments on multiple benchmarks show that FLR consistently outperforms strong baselines while improving robustness and interpretability. Our data and code are available at https://github.com/ToAdventure/FLR.

Multimodal Retrieval-Augmented Generation (MRAG) addresses key limitations of Multimodal Large Language Models (MLLMs), such as hallucination and outdated knowledge. However, current MRAG systems struggle to distinguish whether retrieved multimodal data truly supports the semantic core of an answer or merely provides superficial relevance. Existing metrics often rely on heuristic position-based confidence, which fails to capture the informational density of multimodal entities. To address this, we propose Multi-modal Evidence Grounding (MEG), a semantic-aware metric that quantifies the contribution of retrieved evidence. Unlike standard confidence measures, MEG utilizes Semantic Certainty Anchoring, which dynamically filters out high-frequency stopwords via Inverse Document Frequency (IDF) to focus strictly on information-bearing tokens. Building on MEG, we introduce MEG-RAG, a framework that trains a multimodal reranker to align retrieved evidence with the semantic anchors of the ground truth. By prioritizing high-value content based on semantic grounding rather than token probability distributions, MEG-RAG improves the accuracy and multimodal consistency of generated outputs. Extensive experiments on the ??2RAG benchmark show that MEG-RAG consistently outperforms strong baselines and demonstrates robust generalization across different teacher models. The data and code are available at here.

Recent advancements in diffusion models have shown promising results in sequential recommendation (SR). Existing approaches predominantly rely on implicit conditional diffusion models, which compress user behaviors into a single representation during the forward diffusion process. While effective to some extent, this oversimplification often leads to the loss of sequential and contextual information, which is critical for understanding user behavior. Moreover, explicit information, such as user-item interactions or sequential patterns, remains underutilized, despite its potential to directly guide the recommendation process and improve precision. However, combining implicit and explicit information is non-trivial, as it requires dynamically integrating these complementary signals while avoiding noise and irrelevant patterns within user behaviors. To address these challenges, we propose Dual Conditional Diffusion Models for Sequential Recommendation (DCRec), which effectively integrates implicit and explicit information by embedding dual conditions into both the forward and reverse diffusion processes. This allows the model to retain valuable sequential and contextual information while leveraging explicit user-item interactions to guide the recommendation process. Specifically, we introduce the Dual Conditional Diffusion Transformer (DCDT), which dynamically integrate both implicit and explicit signals throughout the diffusion stages, ensuring contextual understanding and minimizing the influence of irrelevant patterns. Extensive experiments on public benchmark datasets demonstrate that DCRec significantly outperforms state-of-the-art methods.

Retrieval-augmented generation (RAG) has become a powerful framework for enhancing large language models in knowledge-intensive and reasoning tasks. However, as reasoning chains deepen or search trees expand, RAG systems often face two persistent failures: evidence forgetting, where retrieved knowledge is not effectively used, and inefficiency, caused by uncontrolled query expansions and redundant retrieval. These issues reveal a critical gap between retrieval and evidence utilization in current RAG architectures. We propose PruneRAG, a confidence-guided query decomposition framework that builds a structured query decomposition tree to perform stable and efficient reasoning. PruneRAG introduces three key mechanisms: adaptive node expansion that regulates tree width and depth, confidence-guided decisions that accept reliable answers and prune uncertain branches, and fine-grained retrieval that extracts entity-level anchors to improve retrieval precision. Together, these components preserve salient evidence throughout multi-hop reasoning while significantly reducing retrieval overhead. To better analyze evidence misuse, we define the Evidence Forgetting Rate as a metric to quantify cases where golden evidence is retrieved but not correctly used. Extensive experiments across various multi-hop QA benchmarks show that PruneRAG achieves superior accuracy and efficiency over state-of-the-art baselines. The code is publicly available.

Users increasingly interact with content across multiple domains, resulting in sequential behaviors marked by frequent and complex transitions. While Cross-Domain Sequential Recommendation (CDSR) models two-domain interactions, Multi-Domain Sequential Recommendation (MDSR) introduces significantly more domain transitions, compounded by challenges such as domain heterogeneity and imbalance. Existing approaches often overlook the intricacies of domain transitions, tend to overfit to dense domains while underfitting sparse ones, and struggle to scale effectively as the number of domains increases. We propose GMFlowRec, an efficient generative framework for MDSR that models domain-aware transition trajectories via Gaussian Mixture Flow Matching. GMFlowRec integrates: (1) a unified dual-masked Transformer to disentangle domain-invariant and domain-specific intents, (2) a Gaussian Mixture flow field to capture diverse behavioral patterns, and (3) a domain-aligned prior to support frequent and sparse transitions. Extensive experiments on JD and Amazon datasets show GMFlowRec achieves up to 44% NDCG@5 improvement over state-of-the-art baselines while maintaining efficiency with a single unified backbone. Our code and data are available here.

Sentiment bias is newly discovered in Recommender Systems (RSs). Critical users and niche items are disadvantaged by such unfair recommendations. To mitigate this bias, we propose a novel approach by counterfactual inference, which is implemented in two stages. Experiment results validate that our model achieves comparable performance in rating prediction, providing better recommendations and effectively mitigating sentiment bias. To the best of our knowledge, this is the first work to employ counterfactual inference on sentiment bias mitigation in RSs.

Retrieval-augmented large language models (LLMs) have been remarkably competent in various NLP tasks. However, it was observed by previous works that retrieval is not always helpful, especially when the LLM is already knowledgable on the query to answer. Motivated by this, Adaptive Retrieval-Augmented Generation (ARAG) studies retrieving only when the knowledge asked by the query is absent in the LLM. Previous works of ARAG either require accessing the pre-training corpus or prompting with additional model inferences. Aiming to avoid such drawbacks, we propose to determine whether the model is knowledgeable on a query via inspecting the (contextualized) pre-trained token embeddings of LLMs. We hypothesize that such embeddings capture rich information on the model’s intrinsic knowledge base, which enables an efficient way of judging the necessity to retrieve from an external corpus. Extensive experiments demonstrate our ARAG approach’s superior performance across various benchmarks.

For question-answering (QA) tasks, in-context learning (ICL) enables language models (LMs) to generate responses without modifying their parameters by leveraging examples provided in the input. However, the effectiveness of ICL heavily depends on the availability of high-quality examples, which are often scarce due to data privacy constraints, annotation costs, and distribution disparities. A natural solution is to utilize examples stored on client devices, but existing approaches either require transmitting model parameters—incurring significant communication overhead—or fail to fully exploit local datasets, limiting their effectiveness. To address these challenges, we propose Federated In-Context Learning (Fed-ICL), a general framework that enhances ICL through an iterative, collaborative process. Fed-ICL progressively refines responses by leveraging multi-round interactions between clients and a central server, improving answer quality without the need to transmit model parameters. We establish theoretical guarantees for the convergence of Fed-ICL and conduct extensive experiments on standard QA benchmarks, demonstrating that our proposed approach achieves strong performance while maintaining low communication costs.

Forecasting multi-step user behavior trajectories requires reasoning over structured preferences across future actions, a challenge overlooked by traditional sequential recommendation. This problem is critical for applications such as personalized commerce and adaptive content delivery, where anticipating a user’s complete action sequence enhances both satisfaction and business outcomes. We identify an essential limitation of existing paradigms: their inability to capture global, listwise dependencies among sequence items. To address this, we formulate User Behavior Trajectory Prediction (UBTP) as a new task setting that explicitly models longterm user preferences. We introduce Listwise Preference Diffusion Optimization (LPDO), a diffusion-based training framework that directly optimizes structured preferences over entire item sequences. LPDO incorporates a Plackett–Luce supervision signal and derives a tight variational lower bound aligned with listwise ranking likelihoods, enabling coherent preference generation across denoising steps and overcoming the independent-token assumption of prior diffusion methods. To rigorously evaluate multi-step prediction quality, we propose the task-specific metric: Sequential Match (SeqMatch), which measures exact trajectory agreement, and adopt Perplexity (PPL), which assesses probabilistic fidelity. Extensive experiments on real-world user behavior benchmarks demonstrate that LPDO consistently outperforms state-of-the-art baselines, establishing a new benchmark for structured preference learning with diffusion models.

The field of generating recommendations within the framework of causal inference has seen a recent surge.This approach enhances insights into the influence of recommendations on user behavior and helps in identifying the underlying factors. Existing research has often leveraged propensity scores to mitigate bias, albeit at the risk of introducing additional variance. Others have explored the use of unbiased data from randomized controlled trials, although this comes with assumptions that may prove challenging in practice. In this paper, we first present the causality-aware interpretation of recommendations and reveal how the underlying exposure mechanism can bias the maximum likelihood estimation (MLE) of observational feedback. Recognizing that confounders may be elusive, we propose a contrastive self-supervised learning to minimize exposure bias, employing inverse propensity scores and expanding the positive sample set. Building on this foundation, we present a novel contrastive counterfactual learning method (CCL) that incorporates three unique positive sampling strategies grounded in estimated exposure probability or random counterfactual samples. Through extensive experiments on two real-world datasets, we demonstrate that our CCL outperforms the state-of-the-art methods.

Sequential recommender systems (SRSs) aim to predict the subsequent items which may interest users via comprehensively modeling users' complex preference embedded in the sequence of user-item interactions. However, most of existing SRSs often model users' single low-level preference based on item ID information while ignoring the high-level preference revealed by item attribute information, such as item category. Furthermore, they often utilize limited sequence context information to predict the next item while overlooking richer inter-item semantic relations. To this end, in this paper, we proposed a novel hierarchical preference modeling framework to substantially model the complex low- and high-level preference dynamics for accurate sequential recommendation. Specifically, in the framework, a novel dual-transformer module and a novel dual contrastive learning scheme have been designed to discriminatively learn users' low- and high-level preference and to effectively enhance both low- and high-level preference learning respectively. In addition, a novel semantics-enhanced context embedding module has been devised to generate more informative context embedding for further improving the recommendation performance. Extensive experiments on six real-world datasets have demonstrated both the superiority of our proposed method over the state-of-the-art ones and the rationality of our design.

Sequential recommendation aims to predict the next item which interests users via modeling their interest in items over time. Most of the existing works on sequential recommendation model users’ dynamic interest in specific items while overlooking users’ static interest revealed by some static attribute information of items, e.g., category, brand. Moreover, existing works often only consider the positive excitation of a user’s historical interactions on his/her next choice on candidate items while ignoring the commonly existing negative excitation, resulting in insufficiently modeling dynamic interest. The overlook of static interest and negative excitation will lead to incomplete interest modeling and thus impedes the recommendation performance. To this end, in this paper, we propose modeling both static interest and negative excitation for dynamic interest to further improve the recommendation performance. Accordingly, we design a novel Static-Dynamic Interest Learning (SDIL) framework featured with a novel Temporal Positive and Negative Excitation Modeling (TPNE) module for accurate sequential recommendation. TPNE is specially designed for comprehensively modeling dynamic interest based on temporal positive and negative excitation learning. Extensive experiments on three real-world datasets show that SDIL can effectively capture both static and dynamic interest and outperforms state-of-the-art baselines.

This paper addresses the fault estimation (FE) and accommodation issues of interconnected systems by using two new concepts namely interconnected separation principle and constrained interconnected separation principle that allow for the separate design not only between diagnostic observer and fault tolerant controller for each subsystem, but also between observer/controller of each subsystem and those of other ones. Sufficient fault recoverability conditions are established, under which both distributed and decentralized FE and accommodation schemes are provided. The new results help to provide a framework for observer-based fault diagnosis and fault tolerant control of interconnected systems, and are further applied to the meta aircraft configuration that consists of multiple aircraft joined together to illustrate their efficiency.

As we all know, the accuracy of GPS become worse indoors due to the satellite signal attenuation. However, people spend most of his or her time in the indoor environment and the location-based service is needed in the indoor environment. Hence, many indoor positioning techniques have been researched to provide location-based service for visitors in public buildings such as museums, galleries, etc. However, the distinguish of human's moving patterns and initial position detection are two difficult problems in Pedestrian Dead Reckoning (PDR) system. Therefore, A smartphone-based indoor positioning system (called Improved Pedestrian Dead Reckoning, IPDR) using fuzzy theory and WLAN mapping algorithm is presented to solve these problems. In our proposed system, we propose a method called fuzzy step length detection which can distinguish five different human's moving patterns and handle various ways of holding the phone. The WLAN mapping algorithm could calculate the initial position. Calibrated position can be obtained from the IPDR position and the mapping position from WLAN map acquired in advance. The performance of our IPDR system is verified via a set of simulations and the result is positive.

In recent decades, indoor positioning techniques have been researched to support automatic guidance for visitors in public buildings such as museums, galleries, etc. The exhibition goods nearby could be introduced by a smartphone after the location of the visitor has been acquired. This paper presents an indoor positioning scheme utilizing smartphones equipped with inertial and magnetic sensors. The positioning method could handle complicated human motion and various ways to hold the phone. This is achieved by applying an improved Pedestrian Dead Reckoning (PDR) algorithm and an error-tolerant magnetic map matching algorithm. The improved PDR algorithm estimates walking distances and direction in real time. As long as the user stops to appreciate showpieces, the improved PDR component will report relevant data to magnetic calibration component. Based on the updated information, the final location of the user could be calculated utilizing the relevant data, the real-time data of magnetic field sensor and the magnetic map acquired in advance. We evaluate this method using experimental measurements in practice.

As a tradeoff of GEO and LEO, MEO satellite system has more acceptable service performance and it is more appropriate to provide global mobile communications. A MEO satellite system model communicating according to time slots is constructed in the paper. Moreover, in order to improve comprehensive performance of the network, a novel routing algorithm applying Time Evolving Graph based on Pairing Heap is proposed. The Time Evolving Graph is employed to analysis the dynamic topology of the network and the Pairing Heap is applied in the Dijkstra algorithm to reduce the time complexity. By contrast, Fibonacci Heap is also used to optimize Dijkstra algorithm. Finally simulation results show that routing algorithm applying Time Evolving Graph based on Pairing Heap can perform better and reduce the time complexity obviously, and at the same time, Pairing Heap works better than Fibonacci Heap when the number of nodes grows bigger.

The accuracy of Global Positioning System (GPS) is often combined with the reliability of Inertial Navigation System (INS) to accomplish navigation. This paper proposes an innovative way to filter and fuse the GPS and INS information. UKF is employed to simulate the information convergence of the dynamic model which maintains better performance in nonlinear system. So we can obtain a fair precise filtering result when both are online. At the same time, the INS data is trained with the result as training target when it is the unique input. This paper raises the idea that Support Vector Machine (SVM) is adopted to train the INS data during GPS outage and the simulated annealing is applied to realize the optimization of the parameters of kernel function and the penalty function in the SVM algorithm. Therefore, the integration navigation could retain almost as precise as the GPS when the GPS is off-line.