Xin Xin

Shen Gao, Zhengliang Shi, Minghang Zhu et al. · pku

Augmenting large language models (LLMs) with external tools has emerged as a promising approach to extending the capability of LLMs. Although some works employ open-source LLMs for the tool learning task, most of them are trained in a controlled environment in which LLMs only learn to execute the human-provided tools. However, selecting proper tools from the large toolset is also a crucial ability for the tool learning model to be applied in real-world applications. Existing methods usually directly employ self-instruction methods to train the model, which ignores differences in tool complexity. In this paper, we propose the Confucius, a novel tool learning framework to train LLM to use complicated tools in real-world scenarios, which contains two main phases: (1) We first propose a multi-stage learning method to teach the LLM to use various tools from an easy-to-difficult curriculum; (2) thenceforth, we propose the Iterative Self-instruct from Introspective Feedback (ISIF) to dynamically construct the dataset to improve the ability to use the complicated tool. Extensive experiments conducted on both controlled and real-world settings demonstrate the superiority of our tool learning framework in the real-world application scenarios compared to both tuning-free (e.g. ChatGPT, Claude) and tuning-based baselines (e.g. GPT4Tools).

Ziqi Zhao, Xinyu Ma, Liu Yang et al.

On-policy self-distillation (OPSD) improves the reasoning performance of large language models (LLMs) by providing dense token-level supervision for on-policy rollouts. However, existing OPSD methods often yield limited gains on in-domain reasoning and generalize poorly to out-of-domain problems. We identify two key causes: conditioning the self-teacher on a verified solution encourages imitation of training-domain reference trajectories rather than error-specific correction, and applying distillation to the full response can overwrite valid reasoning prefixes and reinforce overfitting. We propose Reflective On-policy Self-Distillation (ROSD), a framework that turns reference-solution imitation into targeted reasoning correction through reflection-guided, error-localized distillation. For each rollout, ROSD uses a self-reflector to extract a corrective idea and locate the first erroneous span. The corrective idea guides the self-teacher toward targeted supervision, while the localized error span restricts distillation to where correction is needed. This design corrects flawed reasoning while preserving valid prefixes. Experiments on multiple in-domain and out-of-domain reasoning benchmarks show that ROSD yields stronger in-domain reasoning performance overall and substantially better out-of-domain generalization than standard OPSD. Code is available at https://github.com/ZiqiZhao1/ROSD.

Qingyao Ai, Ting Bai, Zhao Cao et al. · pku, tsinghua

The research field of Information Retrieval (IR) has evolved significantly, expanding beyond traditional search to meet diverse user information needs. Recently, Large Language Models (LLMs) have demonstrated exceptional capabilities in text understanding, generation, and knowledge inference, opening up exciting avenues for IR research. LLMs not only facilitate generative retrieval but also offer improved solutions for user understanding, model evaluation, and user-system interactions. More importantly, the synergistic relationship among IR models, LLMs, and humans forms a new technical paradigm that is more powerful for information seeking. IR models provide real-time and relevant information, LLMs contribute internal knowledge, and humans play a central role of demanders and evaluators to the reliability of information services. Nevertheless, significant challenges exist, including computational costs, credibility concerns, domain-specific limitations, and ethical considerations. To thoroughly discuss the transformative impact of LLMs on IR research, the Chinese IR community conducted a strategic workshop in April 2023, yielding valuable insights. This paper provides a summary of the workshop's outcomes, including the rethinking of IR's core values, the mutual enhancement of LLMs and IR, the proposal of a novel IR technical paradigm, and open challenges.

Zhen Zhang, Zihan Wang, Xinyu Ma et al.

Generative retrieval (GR) reformulates the Information Retrieval (IR) task as the generation of document identifiers (docIDs). Despite its promise, existing GR models exhibit poor generalization to newly added documents, often failing to generate the correct docIDs. While incremental training offers a straightforward remedy, it is computationally expensive, resource-intensive, and prone to catastrophic forgetting, thereby limiting the scalability and practicality of GR. In this paper, we identify the core bottleneck as the decoder's ability to map hidden states to the correct docIDs of newly added documents. Model editing, which enables targeted parameter modifications for docID mapping, represents a promising solution. However, applying model editing to current GR models is not trivial, which is severely hindered by indistinguishable edit vectors across queries, due to the high overlap of shared docIDs in retrieval results. To address this, we propose DOME (docID-oriented model editing), a novel method that effectively and efficiently adapts GR models to unseen documents. DOME comprises three stages: (1) identification of critical layers, (2) optimization of edit vectors, and (3) construction and application of updates. At its core, DOME employs a hybrid-label adaptive training strategy that learns discriminative edit vectors by combining soft labels, which preserve query-specific semantics for distinguishable updates, with hard labels that enforce precise mapping modifications. Experiments on widely used benchmarks, including NQ and MS MARCO, show that our method significantly improves retrieval performance on new documents while maintaining effectiveness on the original collection. Moreover, DOME achieves this with only about 60% of the training time required by incremental training, considerably reducing computational cost and enabling efficient, frequent model updates.

Ziqi Zhao, Zhaochun Ren, Jiahong Zou et al.

Reinforcement learning with verifiable rewards (RLVR) has proven effective in enhancing the reasoning of large language models (LLMs). Monte Carlo Tree Search (MCTS)-based extensions improve upon vanilla RLVR (e.g., GRPO) by providing tree-based reasoning rollouts that enable fine-grained and segment-level credit assignment. However, existing methods still suffer from limited exploration diversity and inefficient reasoning. To address the above challenges, we propose reinforced efficient reasoning via semantically diverse explorations, i.e., ROSE, for LLMs. To encourage more diverse reasoning exploration, our method incorporates a semantic-entropy-based branching strategy and an $\varepsilon$-exploration mechanism. The former operates on already sampled reasoning rollouts to capture semantic uncertainty and select branching points with high semantic divergence to generate new successive reasoning paths, whereas the latter stochastically initiates reasoning rollouts from the root, preventing the search process from becoming overly local. To improve efficiency, we design a length-aware segment-level advantage estimator that rewards concise and correct reasoning while penalizing unnecessarily long reasoning chains. Extensive experiments on various mathematical reasoning benchmarks with Qwen and Llama models validate the effectiveness and efficiency of ROSE. Codes are available at https://github.com/ZiqiZhao1/ROSE-rl.

Xiaoyu Zhang, Xin Xin, Dongdong Li et al.

Conversational recommender systems (CRSs) often utilize external knowledge graphs (KGs) to introduce rich semantic information and recommend relevant items through natural language dialogues. However, original KGs employed in existing CRSs are often incomplete and sparse, which limits the reasoning capability in recommendation. Moreover, only few of existing studies exploit the dialogue context to dynamically refine knowledge from KGs for better recommendation. To address the above issues, we propose the Variational Reasoning over Incomplete KGs Conversational Recommender (VRICR). Our key idea is to incorporate the large dialogue corpus naturally accompanied with CRSs to enhance the incomplete KGs; and perform dynamic knowledge reasoning conditioned on the dialogue context. Specifically, we denote the dialogue-specific subgraphs of KGs as latent variables with categorical priors for adaptive knowledge graphs refactor. We propose a variational Bayesian method to approximate posterior distributions over dialogue-specific subgraphs, which not only leverages the dialogue corpus for restructuring missing entity relations but also dynamically selects knowledge based on the dialogue context. Finally, we infuse the dialogue-specific subgraphs to decode the recommendation and responses. We conduct experiments on two benchmark CRSs datasets. Experimental results confirm the effectiveness of our proposed method.

Xiaoyu Zhang, Ruobing Xie, Yougang Lyu et al.

Conversational recommender systems (CRSs) are able to elicit user preferences through multi-turn dialogues. They typically incorporate external knowledge and pre-trained language models to capture the dialogue context. Most CRS approaches, trained on benchmark datasets, assume that the standard items and responses in these benchmarks are optimal. However, they overlook that users may express negative emotions with the standard items and may not feel emotionally engaged by the standard responses. This issue leads to a tendency to replicate the logic of recommenders in the dataset instead of aligning with user needs. To remedy this misalignment, we introduce empathy within a CRS. With empathy we refer to a system's ability to capture and express emotions. We propose an empathetic conversational recommender (ECR) framework. ECR contains two main modules: emotion-aware item recommendation and emotion-aligned response generation. Specifically, we employ user emotions to refine user preference modeling for accurate recommendations. To generate human-like emotional responses, ECR applies retrieval-augmented prompts to fine-tune a pre-trained language model aligning with emotions and mitigating hallucination. To address the challenge of insufficient supervision labels, we enlarge our empathetic data using emotion labels annotated by large language models and emotional reviews collected from external resources. We propose novel evaluation metrics to capture user satisfaction in real-world CRS scenarios. Our experiments on the ReDial dataset validate the efficacy of our framework in enhancing recommendation accuracy and improving user satisfaction.

Yu Wang, Xin Xin, Zaiqiao Meng et al.

Learning from corrupted labels is very common in real-world machine-learning applications. Memorizing such noisy labels could affect the learning of the model, leading to sub-optimal performances. In this work, we propose a novel framework to learn robust machine-learning models from noisy labels. Through an empirical study, we find that different models make relatively similar predictions on clean examples, while the predictions on noisy examples vary much more across different models. Motivated by this observation, we propose \em denoising with cross-model agreement \em (DeCA) which aims to minimize the KL-divergence between the true label distributions parameterized by two machine learning models while maximizing the likelihood of data observation. We employ the proposed DeCA on both the binary label scenario and the multiple label scenario. For the binary label scenario, we select implicit feedback recommendation as the downstream task and conduct experiments with four state-of-the-art recommendation models on four datasets. For the multiple-label scenario, the downstream application is image classification on two benchmark datasets. Experimental results demonstrate that the proposed methods significantly improve the model performance compared with normal training and other denoising methods on both binary and multiple-label scenarios.

Zhiwei Xu, Hangyu Mao, Nianmin Zhang et al.

In partially observable multi-agent systems, agents typically only have access to local observations. This severely hinders their ability to make precise decisions, particularly during decentralized execution. To alleviate this problem and inspired by image outpainting, we propose State Inference with Diffusion Models (SIDIFF), which uses diffusion models to reconstruct the original global state based solely on local observations. SIDIFF consists of a state generator and a state extractor, which allow agents to choose suitable actions by considering both the reconstructed global state and local observations. In addition, SIDIFF can be effortlessly incorporated into current multi-agent reinforcement learning algorithms to improve their performance. Finally, we evaluated SIDIFF on different experimental platforms, including Multi-Agent Battle City (MABC), a novel and flexible multi-agent reinforcement learning environment we developed. SIDIFF achieved desirable results and outperformed other popular algorithms.

Junchen Fu, Xuri Ge, Xin Xin et al.

Multimodal representation learning has attracted increasing attention in AI, driven by the strong performance of large, pretrained multimodal foundation models such as Qwen, LLaVA, and CLIP. These models deliver impressive performance on a range of multimodal information retrieval (MIR) tasks, including web search, cross-modal retrieval, and recommender systems. Yet their massive parameter counts create major efficiency bottlenecks when adapting their representations for IR tasks during training, deployment, and inference. These limitations hinder the practical use of foundation models for representation learning in information retrieval. To address these issues, we propose organizing the EReL@MIR workshop at MM 2026, bringing together researchers from academia and industry to discuss emerging solutions, open challenges, and new efficiency metrics and benchmarks for multimodal IR representation learning in the foundation-model era. The workshop's official website is available at https://erel-mir.github.io/.

Fangan Dong, Zuming Yan, Xuri Ge et al.

Despite the strong reasoning capabilities of recent large language models (LLMs), achieving reliable performance on challenging tasks often requires post-training or computationally expensive sampling strategies, limiting their practical efficiency. In this work, we first show that a small subset of neurons in LLMs exhibits strong predictive correlations with reasoning correctness. Based on this observation, we propose AdaRAS (Adaptive Reasoning Activation Steering), a lightweight test-time framework that improves reasoning reliability by selectively intervening on neuron activations. AdaRAS identifies Reasoning-Critical Neurons (RCNs) via a polarity-aware mean-difference criterion and adaptively steers their activations during inference, enhancing incorrect reasoning traces while avoiding degradation on already-correct cases. Experiments on 10 mathematics and coding benchmarks demonstrate consistent improvements, including over 13% gains on AIME-24 and AIME-25. Moreover, AdaRAS exhibits strong transferability across datasets and scalability to stronger models, outperforming post-training methods without additional training or sampling cost.

Junchen Fu, Xuri Ge, Xin Xin et al.

Multimodal foundation models are transformative in sequential recommender systems, leveraging powerful representation learning capabilities. While Parameter-efficient Fine-tuning (PEFT) is commonly used to adapt foundation models for recommendation tasks, most research prioritizes parameter efficiency, often overlooking critical factors like GPU memory efficiency and training speed. Addressing this gap, our paper introduces IISAN (Intra- and Inter-modal Side Adapted Network for Multimodal Representation), a simple plug-and-play architecture using a Decoupled PEFT structure and exploiting both intra- and inter-modal adaptation. IISAN matches the performance of full fine-tuning (FFT) and state-of-the-art PEFT. More importantly, it significantly reduces GPU memory usage - from 47GB to just 3GB for multimodal sequential recommendation tasks. Additionally, it accelerates training time per epoch from 443s to 22s compared to FFT. This is also a notable improvement over the Adapter and LoRA, which require 37-39 GB GPU memory and 350-380 seconds per epoch for training. Furthermore, we propose a new composite efficiency metric, TPME (Training-time, Parameter, and GPU Memory Efficiency) to alleviate the prevalent misconception that "parameter efficiency represents overall efficiency". TPME provides more comprehensive insights into practical efficiency comparisons between different methods. Besides, we give an accessible efficiency analysis of all PEFT and FFT approaches, which demonstrate the superiority of IISAN. We release our codes and other materials at https://github.com/GAIR-Lab/IISAN.

Weiqin Yang, Jiawei Chen, Xin Xin et al.

Softmax Loss (SL) is widely applied in recommender systems (RS) and has demonstrated effectiveness. This work analyzes SL from a pairwise perspective, revealing two significant limitations: 1) the relationship between SL and conventional ranking metrics like DCG is not sufficiently tight; 2) SL is highly sensitive to false negative instances. Our analysis indicates that these limitations are primarily due to the use of the exponential function. To address these issues, this work extends SL to a new family of loss functions, termed Pairwise Softmax Loss (PSL), which replaces the exponential function in SL with other appropriate activation functions. While the revision is minimal, we highlight three merits of PSL: 1) it serves as a tighter surrogate for DCG with suitable activation functions; 2) it better balances data contributions; and 3) it acts as a specific BPR loss enhanced by Distributionally Robust Optimization (DRO). We further validate the effectiveness and robustness of PSL through empirical experiments. The code is available at https://github.com/Tiny-Snow/IR-Benchmark.

Junchen Fu, Xuri Ge, Xin Xin et al.

Multimodal foundation models (MFMs) have revolutionized sequential recommender systems through advanced representation learning. While Parameter-efficient Fine-tuning (PEFT) is commonly used to adapt these models, studies often prioritize parameter efficiency, neglecting GPU memory and training speed. To address this, we introduced the IISAN framework, significantly enhancing efficiency. However, IISAN was limited to symmetrical MFMs and identical text and image encoders, preventing the use of state-of-the-art Large Language Models. To overcome this, we developed IISAN-Versa, a versatile plug-and-play architecture compatible with both symmetrical and asymmetrical MFMs. IISAN-Versa employs a Decoupled PEFT structure and utilizes both intra- and inter-modal adaptation. It effectively handles asymmetry through a simple yet effective combination of group layer-dropping and dimension transformation alignment. Our research demonstrates that IISAN-Versa effectively adapts large text encoders, and we further identify a scaling effect where larger encoders generally perform better. IISAN-Versa also demonstrates strong versatility in our defined multimodal scenarios, which include raw titles and captions generated from images and videos. Additionally, IISAN-Versa achieved state-of-the-art performance on the Microlens public benchmark. We release our code at https://github.com/GAIR-Lab/IISAN.

Ziqi Zhao, Zhaochun Ren, Liu Yang et al.

Offline reinforcement learning (RL) aims to learn policies without online explorations. To enlarge the training data, model-based offline RL learns a dynamics model which is utilized as a virtual environment to generate simulation data and enhance policy learning. However, existing data augmentation methods for offline RL suffer from (i) trivial improvement from short-horizon simulation; and (ii) the lack of evaluation and correction for generated data, leading to low-qualified augmentation. In this paper, we propose offline trajectory optimization for offline reinforcement learning (OTTO). The key motivation is to conduct long-horizon simulation and then utilize model uncertainty to evaluate and correct the augmented data. Specifically, we propose an ensemble of Transformers, a.k.a. World Transformers, to predict environment state dynamics and the reward function. Three strategies are proposed to use World Transformers to generate long-horizon trajectory simulation by perturbing the actions in the offline data. Then, an uncertainty-based World Evaluator is introduced to firstly evaluate the confidence of the generated trajectories and then perform the correction for low-confidence data. Finally, we jointly use the original data with the corrected augmentation data to train an offline RL algorithm. OTTO serves as a plug-in module and can be integrated with existing model-free offline RL methods. Experiments on various benchmarks show that OTTO can effectively improve the performance of representative offline RL algorithms, including in complex environments with sparse rewards like AntMaze. Codes are available at https://github.com/ZiqiZhao1/OTTO.

Jitai Hao, WeiWei Sun, Xin Xin et al.

Parameter-Efficient Fine-tuning (PEFT) facilitates the fine-tuning of Large Language Models (LLMs) under limited resources. However, the fine-tuning performance with PEFT on complex, knowledge-intensive tasks is limited due to the constrained model capacity, which originates from the limited number of additional trainable parameters. To overcome this limitation, we introduce a novel mechanism that fine-tunes LLMs with adapters of larger size yet memory-efficient. This is achieved by leveraging the inherent activation sparsity in the Feed-Forward Networks (FFNs) of LLMs and utilizing the larger capacity of Central Processing Unit (CPU) memory compared to Graphics Processing Unit (GPU). We store and update the parameters of larger adapters on the CPU. Moreover, we employ a Mixture of Experts (MoE)-like architecture to mitigate unnecessary CPU computations and reduce the communication volume between the GPU and CPU. This is particularly beneficial over the limited bandwidth of PCI Express (PCIe). Our method can achieve fine-tuning results comparable to those obtained with larger memory capacities, even when operating under more limited resources such as a 24GB memory single GPU setup, with acceptable loss in training efficiency. Our codes are available at https://github.com/CURRENTF/MEFT.

Xin Xin, Alexandros Karatzoglou, Ioannis Arapakis et al.

Casting session-based or sequential recommendation as reinforcement learning (RL) through reward signals is a promising research direction towards recommender systems (RS) that maximize cumulative profits. However, the direct use of RL algorithms in the RS setting is impractical due to challenges like off-policy training, huge action spaces and lack of sufficient reward signals. Recent RL approaches for RS attempt to tackle these challenges by combining RL and (self-)supervised sequential learning, but still suffer from certain limitations. For example, the estimation of Q-values tends to be biased toward positive values due to the lack of negative reward signals. Moreover, the Q-values also depend heavily on the specific timestamp of a sequence. To address the above problems, we propose negative sampling strategy for training the RL component and combine it with supervised sequential learning. We call this method Supervised Negative Q-learning (SNQN). Based on sampled (negative) actions (items), we can calculate the "advantage" of a positive action over the average case, which can be further utilized as a normalized weight for learning the supervised sequential part. This leads to another learning framework: Supervised Advantage Actor-Critic (SA2C). We instantiate SNQN and SA2C with four state-of-the-art sequential recommendation models and conduct experiments on two real-world datasets. Experimental results show that the proposed approaches achieve significantly better performance than state-of-the-art supervised methods and existing self-supervised RL methods . Code will be open-sourced.

Yu Wang, Xin Xin, Zaiqiao Meng et al.

Learning from implicit feedback is one of the most common cases in the application of recommender systems. Generally speaking, interacted examples are considered as positive while negative examples are sampled from uninteracted ones. However, noisy examples are prevalent in real-world implicit feedback. A noisy positive example could be interacted but it actually leads to negative user preference. A noisy negative example which is uninteracted because of unawareness of the user could also denote potential positive user preference. Conventional training methods overlook these noisy examples, leading to sub-optimal recommendations. In this work, we propose a novel framework to learn robust recommenders from implicit feedback. Through an empirical study, we find that different models make relatively similar predictions on clean examples which denote the real user preference, while the predictions on noisy examples vary much more across different models. Motivated by this observation, we propose denoising with cross-model agreement(DeCA) which aims to minimize the KL-divergence between the real user preference distributions parameterized by two recommendation models while maximizing the likelihood of data observation. We employ the proposed DeCA on four state-of-the-art recommendation models and conduct experiments on four datasets. Experimental results demonstrate that DeCA significantly improves recommendation performance compared with normal training and other denoising methods. Codes will be open-sourced.

Jiawei Chen, Hande Dong, Yang Qiu et al.

Recommender systems rely on user behavior data like ratings and clicks to build personalization model. However, the collected data is observational rather than experimental, causing various biases in the data which significantly affect the learned model. Most existing work for recommendation debiasing, such as the inverse propensity scoring and imputation approaches, focuses on one or two specific biases, lacking the universal capacity that can account for mixed or even unknown biases in the data. Towards this research gap, we first analyze the origin of biases from the perspective of \textit{risk discrepancy} that represents the difference between the expectation empirical risk and the true risk. Remarkably, we derive a general learning framework that well summarizes most existing debiasing strategies by specifying some parameters of the general framework. This provides a valuable opportunity to develop a universal solution for debiasing, e.g., by learning the debiasing parameters from data. However, the training data lacks important signal of how the data is biased and what the unbiased data looks like. To move this idea forward, we propose \textit{AotoDebias} that leverages another (small) set of uniform data to optimize the debiasing parameters by solving the bi-level optimization problem with meta-learning. Through theoretical analyses, we derive the generalization bound for AutoDebias and prove its ability to acquire the appropriate debiasing strategy. Extensive experiments on two real datasets and a simulated dataset demonstrated effectiveness of AutoDebias. The code is available at \url{https://github.com/DongHande/AutoDebias}.

Xin Xin, Liu Yang, Ziqi Zhao et al.

Session-based recommendation predicts users' future interests from previous interactions in a session. Despite the memorizing of historical samples, the request of unlearning, i.e., to remove the effect of certain training samples, also occurs for reasons such as user privacy or model fidelity. However, existing studies on unlearning are not tailored for the session-based recommendation. On the one hand, these approaches cannot achieve satisfying unlearning effects due to the collaborative correlations and sequential connections between the unlearning item and the remaining items in the session. On the other hand, seldom work has conducted the research to verify the unlearning effectiveness in the session-based recommendation scenario. In this paper, we propose SRU, a session-based recommendation unlearning framework, which enables high unlearning efficiency, accurate recommendation performance, and improved unlearning effectiveness in session-based recommendation. Specifically, we first partition the training sessions into separate sub-models according to the similarity across the sessions, then we utilize an attention-based aggregation layer to fuse the hidden states according to the correlations between the session and the centroid of the data in the sub-model. To improve the unlearning effectiveness, we further propose three extra data deletion strategies, including collaborative extra deletion (CED), neighbor extra deletion (NED), and random extra deletion (RED). Besides, we propose an evaluation metric that measures whether the unlearning sample can be inferred after the data deletion to verify the unlearning effectiveness. We implement SRU with three representative session-based recommendation models and conduct experiments on three benchmark datasets. Experimental results demonstrate the effectiveness of our methods.

Zhen Zhang, Jujia Zhao, Xinyu Ma et al.

Cold-start recommendation remains a central challenge in dynamic, open-world platforms, requiring models to recommend for newly registered users (user cold-start) and to recommend newly introduced items to existing users (item cold-start) under sparse or missing interaction signals. Recent generative recommenders built on pre-trained language models (PLMs) are often expected to mitigate cold-start by using item semantic information (e.g., titles and descriptions) and test-time conditioning on limited user context. However, cold-start is rarely treated as a primary evaluation setting in existing studies, and reported gains are difficult to interpret because key design choices, such as model scale, identifier design, and training strategy, are frequently changed together. In this work, we present a systematic reproducibility study of generative recommendation under a unified suite of cold-start protocols.

Junchen Fu, Xuri Ge, Kaiwen Zheng et al.

Popular Micro-videos, dominant on platforms like TikTok and YouTube, hold significant commercial value. The rise of high-quality AI-generated content has spurred interest in AI-driven micro-video creation. However, despite the advanced capabilities of large language models (LLMs) like ChatGPT and DeepSeek in text generation and reasoning, their potential to assist the creation of popular micro-videos remains largely unexplored. In this paper, we conduct an empirical study on LLM-assisted popular micro-video generation (LLMPopcorn). Specifically, we investigate the following research questions: (i) How can LLMs be effectively utilized to assist popular micro-video generation? (ii) To what extent can prompt-based enhancements optimize the LLM-generated content for higher popularity? (iii) How well do various LLMs and video generators perform in the popular micro-video generation task? By exploring these questions, we show that advanced LLMs like DeepSeek-V3 enable micro-video generation to achieve popularity comparable to human-created content. Prompt enhancements further boost popularity, and benchmarking highlights DeepSeek-V3 and DeepSeek-R1 among LLMs, while LTX-Video and HunyuanVideo lead in video generation. This pioneering work advances AI-assisted micro-video creation, uncovering new research opportunities. We will release the code and datasets to support future studies.

Xuri Ge, Chunhao Wang, Xindi Wang et al.

Composed Image Retrieval (CIR) aims to retrieve target images based on a reference image and modified texts. However, existing methods often struggle to extract the correct semantic cues from the reference image that best reflect the user's intent under textual modification prompts, resulting in interference from irrelevant visual noise. In this paper, we propose a novel Multi-level Vision Selection by Multi-modal Chain-of-Thought Reasoning (MCoT-MVS) for CIR, integrating attention-aware multi-level vision features guided by reasoning cues from a multi-modal large language model (MLLM). Specifically, we leverage an MLLM to perform chain-of-thought reasoning on the multimodal composed input, generating the retained, removed, and target-inferred texts. These textual cues subsequently guide two reference visual attention selection modules to selectively extract discriminative patch-level and instance-level semantics from the reference image. Finally, to effectively fuse these multi-granular visual cues with the modified text and the imagined target description, we design a weighted hierarchical combination module to align the composed query with target images in a unified embedding space. Extensive experiments on two CIR benchmarks, namely CIRR and FashionIQ, demonstrate that our approach consistently outperforms existing methods and achieves new state-of-the-art performance. Code and trained models are publicly released.

Zhengfei Xu, Sijia Zhao, Yanchao Hao et al.

Visual Entity Linking (VEL) is a crucial task for achieving fine-grained visual understanding, matching objects within images (visual mentions) to entities in a knowledge base. Previous VEL tasks rely on textual inputs, but writing queries for complex scenes can be challenging. Visual inputs like clicks or bounding boxes offer a more convenient alternative. Therefore, we propose a new task, Pixel-Level Visual Entity Linking (PL-VEL), which uses pixel masks from visual inputs to refer to objects, supplementing reference methods for VEL. To facilitate research on this task, we have constructed the MaskOVEN-Wiki dataset through an entirely automatic reverse region-entity annotation framework. This dataset contains over 5 million annotations aligning pixel-level regions with entity-level labels, which will advance visual understanding towards fine-grained. Moreover, as pixel masks correspond to semantic regions in an image, we enhance previous patch-interacted attention with region-interacted attention by a visual semantic tokenization approach. Manual evaluation results indicate that the reverse annotation framework achieved a 94.8% annotation success rate. Experimental results show that models trained on this dataset improved accuracy by 18 points compared to zero-shot models. Additionally, the semantic tokenization method achieved a 5-point accuracy improvement over the trained baseline.

Dusan Stamenkovic, Alexandros Karatzoglou, Ioannis Arapakis et al.

Since the inception of Recommender Systems (RS), the accuracy of the recommendations in terms of relevance has been the golden criterion for evaluating the quality of RS algorithms. However, by focusing on item relevance, one pays a significant price in terms of other important metrics: users get stuck in a "filter bubble" and their array of options is significantly reduced, hence degrading the quality of the user experience and leading to churn. Recommendation, and in particular session-based/sequential recommendation, is a complex task with multiple - and often conflicting objectives - that existing state-of-the-art approaches fail to address. In this work, we take on the aforementioned challenge and introduce Scalarized Multi-Objective Reinforcement Learning (SMORL) for the RS setting, a novel Reinforcement Learning (RL) framework that can effectively address multi-objective recommendation tasks. The proposed SMORL agent augments standard recommendation models with additional RL layers that enforce it to simultaneously satisfy three principal objectives: accuracy, diversity, and novelty of recommendations. We integrate this framework with four state-of-the-art session-based recommendation models and compare it with a single-objective RL agent that only focuses on accuracy. Our experimental results on two real-world datasets reveal a substantial increase in aggregate diversity, a moderate increase in accuracy, reduced repetitiveness of recommendations, and demonstrate the importance of reinforcing diversity and novelty as complementary objectives.

Yunzhe Li, Yue Ding, Bo Chen et al.

In collaborative filtering, it is an important way to make full use of social information to improve the recommendation quality, which has been proved to be effective because user behavior will be affected by her friends. However, existing works leverage the social relationship to aggregate user features from friends' historical behavior sequences in a user-level indirect paradigm. A significant defect of the indirect paradigm is that it ignores the temporal relationships between behavior events across users. In this paper, we propose a novel time-aware sequential recommendation framework called Social Temporal Excitation Networks (STEN), which introduces temporal point processes to model the fine-grained impact of friends' behaviors on the user s dynamic interests in an event-level direct paradigm. Moreover, we propose to decompose the temporal effect in sequential recommendation into social mutual temporal effect and ego temporal effect. Specifically, we employ a social heterogeneous graph embedding layer to refine user representation via structural information. To enhance temporal information propagation, STEN directly extracts the fine-grained temporal mutual influence of friends' behaviors through the mutually exciting temporal network. Besides, the user s dynamic interests are captured through the self-exciting temporal network. Extensive experiments on three real-world datasets show that STEN outperforms state-of-the-art baseline methods. Moreover, STEN provides event-level recommendation explainability, which is also illustrated experimentally.

Yule Wang, Xin Xin, Yue Ding et al.

Recommender system based on historical user-item interactions is of vital importance for web-based services. However, the observed data used to train the recommender model suffers from severe bias issues. Practically, the item frequency distribution of the dataset is a highly skewed power-law distribution. Interactions of a small fraction of head items account for almost the whole training data. The normal training paradigm from such biased data tends to repetitively generate recommendations from the head items, which further exacerbates the biases and affects the exploration of potentially interesting items from the niche set. In this work, we innovatively explore the central theme of recommendation debiasing from an item cluster-wise multi-objective optimization perspective. Aiming to balance the learning on various item clusters that differ in popularity during the training process, we propose a model-agnostic framework namely Item Cluster-Wise Pareto-Efficient Recommendation (ICPE). In detail, we define our item cluster-wise optimization target as the recommender model should balance all item clusters that differ in popularity, thus we set the model learning on each item cluster as a unique optimization objective. To achieve this goal, we first explore items' popularity levels from a novel causal reasoning perspective. Then, we devise popularity discrepancy-based bisecting clustering to separate the item clusters. Next, we adaptively find the overall harmonious gradient direction for cluster-wise optimization objectives from a Pareto-efficient solver. Finally, in the prediction stage, we perform counterfactual inference to further eliminate the impact of global propensity. Extensive experimental results verify the superiorities of ICPE on overall recommendation performance and biases elimination.

Guojun Yan, Jiahuan Pei, Pengjie Ren et al.

Medical dialogue systems (MDSs) aim to assist doctors and patients with a range of professional medical services, i.e., diagnosis, treatment and consultation. The development of MDSs is hindered because of a lack of resources. In particular. (1) there is no dataset with large-scale medical dialogues that covers multiple medical services and contains fine-grained medical labels (i.e., intents, actions, slots, values), and (2) there is no set of established benchmarks for MDSs for multi-domain, multi-service medical dialogues. In this paper, we present ReMeDi, a set of resource for medical dialogues. ReMeDi consists of two parts, the ReMeDi dataset and the ReMeDi benchmarks. The ReMeDi dataset contains 96,965 conversations between doctors and patients, including 1,557 conversations with fine-gained labels. It covers 843 types of diseases, 5,228 medical entities, and 3 specialties of medical services across 40 domains. To the best of our knowledge, the ReMeDi dataset is the only medical dialogue dataset that covers multiple domains and services, and has fine-grained medical labels. The second part of the ReMeDi resources consists of a set of state-of-the-art models for (medical) dialogue generation. The ReMeDi benchmark has the following methods: (1) pretrained models (i.e., BERT-WWM, BERT-MED, GPT2, and MT5) trained, validated, and tested on the ReMeDi dataset, and (2) a self-supervised contrastive learning(SCL) method to expand the ReMeDi dataset and enhance the training of the state-of-the-art pretrained models. We describe the creation of the ReMeDi dataset, the ReMeDi benchmarking methods, and establish experimental results using the ReMeDi benchmarking methods on the ReMeDi dataset for future research to compare against. With this paper, we share the dataset, implementations of the benchmarks, and evaluation scripts.

Paula Gómez Duran, Alexandros Karatzoglou, Jordi Vitrià et al.

Modern recommender systems (RS) work by processing a number of signals that can be inferred from large sets of user-item interaction data. The main signal to analyze stems from the raw matrix that represents interactions. However, we can increase the performance of RS by considering other kinds of signals like the context of interactions, which could be, for example, the time or date of the interaction, the user location, or sequential data corresponding to the historical interactions of the user with the system. These complex, context-based interaction signals are characterized by a rich relational structure that can be represented by a multi-partite graph. Graph Convolutional Networks (GCNs) have been used successfully in collaborative filtering with simple user-item interaction data. In this work, we generalize the use of GCNs for N-partite graphs by considering N multiple context dimensions and propose a simple way for their seamless integration in modern deep learning RS architectures. More specifically, we define a graph convolutional embedding layer for N-partite graphs that processes user-item-context interactions, and constructs node embeddings by leveraging their relational structure. Experiments on several datasets from recommender systems to drug re-purposing show the benefits of the introduced GCN embedding layer by measuring the performance of different context-enriched tasks.

Fuli Feng, Weiran Huang, Xiangnan He et al.

Graph Convolutional Network (GCN) is an emerging technique for information retrieval (IR) applications. While GCN assumes the homophily property of a graph, real-world graphs are never perfect: the local structure of a node may contain discrepancy, e.g., the labels of a node's neighbors could vary. This pushes us to consider the discrepancy of local structure in GCN modeling. Existing work approaches this issue by introducing an additional module such as graph attention, which is expected to learn the contribution of each neighbor. However, such module may not work reliably as expected, especially when there lacks supervision signal, e.g., when the labeled data is small. Moreover, existing methods focus on modeling the nodes in the training data, and never consider the local structure discrepancy of testing nodes. This work focuses on the local structure discrepancy issue for testing nodes, which has received little scrutiny. From a novel perspective of causality, we investigate whether a GCN should trust the local structure of a testing node when predicting its label. To this end, we analyze the working mechanism of GCN with causal graph, estimating the causal effect of a node's local structure for the prediction. The idea is simple yet effective: given a trained GCN model, we first intervene the prediction by blocking the graph structure; we then compare the original prediction with the intervened prediction to assess the causal effect of the local structure on the prediction. Through this way, we can eliminate the impact of local structure discrepancy and make more accurate prediction. Extensive experiments on seven node classification datasets show that our method effectively enhances the inference stage of GCN.

Xin Xin, Alexandros Karatzoglou, Ioannis Arapakis et al.

In session-based or sequential recommendation, it is important to consider a number of factors like long-term user engagement, multiple types of user-item interactions such as clicks, purchases etc. The current state-of-the-art supervised approaches fail to model them appropriately. Casting sequential recommendation task as a reinforcement learning (RL) problem is a promising direction. A major component of RL approaches is to train the agent through interactions with the environment. However, it is often problematic to train a recommender in an on-line fashion due to the requirement to expose users to irrelevant recommendations. As a result, learning the policy from logged implicit feedback is of vital importance, which is challenging due to the pure off-policy setting and lack of negative rewards (feedback). In this paper, we propose self-supervised reinforcement learning for sequential recommendation tasks. Our approach augments standard recommendation models with two output layers: one for self-supervised learning and the other for RL. The RL part acts as a regularizer to drive the supervised layer focusing on specific rewards(e.g., recommending items which may lead to purchases rather than clicks) while the self-supervised layer with cross-entropy loss provides strong gradient signals for parameter updates. Based on such an approach, we propose two frameworks namely Self-Supervised Q-learning(SQN) and Self-Supervised Actor-Critic(SAC). We integrate the proposed frameworks with four state-of-the-art recommendation models. Experimental results on two real-world datasets demonstrate the effectiveness of our approach.

Xin Xin, Alexandros Karatzoglou, Ioannis Arapakis et al.

Graph Convolution Networks (GCN) are widely used in learning graph representations due to their effectiveness and efficiency. However, they suffer from the notorious over-smoothing problem, in which the learned representations of densely connected nodes converge to alike vectors when many (>3) graph convolutional layers are stacked. In this paper, we argue that there-normalization trick used in GCN leads to overly homogeneous information propagation, which is the source of over-smoothing. To address this problem, we propose Graph Highway Networks(GHNet) which utilize gating units to automatically balance the trade-off between homogeneity and heterogeneity in the GCN learning process. The gating units serve as direct highways to maintain heterogeneous information from the node itself after feature propagation. This design enables GHNet to achieve much larger receptive fields per node without over-smoothing and thus access to more of the graph connectivity information. Experimental results on benchmark datasets demonstrate the superior performance of GHNet over GCN and related models.

Enneng Yang, Xin Xin, Li Shen et al.

Factorization machine (FM) is an effective model for feature-based recommendation which utilizes inner product to capture second-order feature interactions. However, one of the major drawbacks of FM is that it couldn't capture complex high-order interaction signals. A common solution is to change the interaction function, such as stacking deep neural networks on the top of FM. In this work, we propose an alternative approach to model high-order interaction signals in the embedding level, namely Generalized Embedding Machine (GEM). The embedding used in GEM encodes not only the information from the feature itself but also the information from other correlated features. Under such situation, the embedding becomes high-order. Then we can incorporate GEM with FM and even its advanced variants to perform feature interactions. More specifically, in this paper we utilize graph convolution networks (GCN) to generate high-order embeddings. We integrate GEM with several FM-based models and conduct extensive experiments on two real-world datasets. The results demonstrate significant improvement of GEM over corresponding baselines.

Xin Xin, Xiangnan He, Yongfeng Zhang et al.

Existing item-based collaborative filtering (ICF) methods leverage only the relation of collaborative similarity. Nevertheless, there exist multiple relations between items in real-world scenarios. Distinct from the collaborative similarity that implies co-interact patterns from the user perspective, these relations reveal fine-grained knowledge on items from different perspectives of meta-data, functionality, etc. However, how to incorporate multiple item relations is less explored in recommendation research. In this work, we propose Relational Collaborative Filtering (RCF), a general framework to exploit multiple relations between items in recommender system. We find that both the relation type and the relation value are crucial in inferring user preference. To this end, we develop a two-level hierarchical attention mechanism to model user preference. The first-level attention discriminates which types of relations are more important, and the second-level attention considers the specific relation values to estimate the contribution of a historical item in recommending the target item. To make the item embeddings be reflective of the relational structure between items, we further formulate a task to preserve the item relations, and jointly train it with the recommendation task of preference modeling. Empirical results on two real datasets demonstrate the strong performance of RCF. Furthermore, we also conduct qualitative analyses to show the benefits of explanations brought by the modeling of multiple item relations.

Ping Guo, Fuqing Duan, Pei Wang et al.

Discovering pulsars is a significant and meaningful research topic in the field of radio astronomy. With the advent of astronomical instruments such as he Five-hundred-meter Aperture Spherical Telescope (FAST) in China, data volumes and data rates are exponentially growing. This fact necessitates a focus on artificial intelligence (AI) technologies that can perform the automatic pulsar candidate identification to mine large astronomical data sets. Automatic pulsar candidate identification can be considered as a task of determining potential candidates for further investigation and eliminating noises of radio frequency interferences or other non-pulsar signals. It is very hard to raise the performance of DCNN-based pulsar identification because the limited training samples restrict network structure to be designed deep enough for learning good features as well as the crucial class imbalance problem due to very limited number of real pulsar samples. To address these problems, we proposed a framework which combines deep convolution generative adversarial network (DCGAN) with support vector machine (SVM) to deal with imbalance class problem and to improve pulsar identification accuracy. DCGAN is used as sample generation and feature learning model, and SVM is adopted as the classifier for predicting candidate's labels in the inference stage. The proposed framework is a novel technique which not only can solve imbalance class problem but also can learn discriminative feature representations of pulsar candidates instead of computing hand-crafted features in preprocessing steps too, which makes it more accurate for automatic pulsar candidate selection. Experiments on two pulsar datasets verify the effectiveness and efficiency of our proposed method.