Junjie Zhang

Wayne Xin Zhao, Kun Zhou, Junyi Li et al.

Language is essentially a complex, intricate system of human expressions governed by grammatical rules. It poses a significant challenge to develop capable AI algorithms for comprehending and grasping a language. As a major approach, language modeling has been widely studied for language understanding and generation in the past two decades, evolving from statistical language models to neural language models. Recently, pre-trained language models (PLMs) have been proposed by pre-training Transformer models over large-scale corpora, showing strong capabilities in solving various NLP tasks. Since researchers have found that model scaling can lead to performance improvement, they further study the scaling effect by increasing the model size to an even larger size. Interestingly, when the parameter scale exceeds a certain level, these enlarged language models not only achieve a significant performance improvement but also show some special abilities that are not present in small-scale language models. To discriminate the difference in parameter scale, the research community has coined the term large language models (LLM) for the PLMs of significant size. Recently, the research on LLMs has been largely advanced by both academia and industry, and a remarkable progress is the launch of ChatGPT, which has attracted widespread attention from society. The technical evolution of LLMs has been making an important impact on the entire AI community, which would revolutionize the way how we develop and use AI algorithms. In this survey, we review the recent advances of LLMs by introducing the background, key findings, and mainstream techniques. In particular, we focus on four major aspects of LLMs, namely pre-training, adaptation tuning, utilization, and capacity evaluation. Besides, we also summarize the available resources for developing LLMs and discuss the remaining issues for future directions.

Wayne Xin Zhao, Kun Zhou, Junyi Li et al.

Language is essentially a complex, intricate system of human expressions governed by grammatical rules. It poses a significant challenge to develop capable AI algorithms for comprehending and grasping a language. As a major approach, language modeling has been widely studied for language understanding and generation in the past two decades, evolving from statistical language models to neural language models. Recently, pre-trained language models (PLMs) have been proposed by pre-training Transformer models over large-scale corpora, showing strong capabilities in solving various NLP tasks. Since researchers have found that model scaling can lead to performance improvement, they further study the scaling effect by increasing the model size to an even larger size. Interestingly, when the parameter scale exceeds a certain level, these enlarged language models not only achieve a significant performance improvement but also show some special abilities that are not present in small-scale language models. To discriminate the difference in parameter scale, the research community has coined the term large language models (LLM) for the PLMs of significant size. Recently, the research on LLMs has been largely advanced by both academia and industry, and a remarkable progress is the launch of ChatGPT, which has attracted widespread attention from society. The technical evolution of LLMs has been making an important impact on the entire AI community, which would revolutionize the way how we develop and use AI algorithms. In this survey, we review the recent advances of LLMs by introducing the background, key findings, and mainstream techniques. In particular, we focus on four major aspects of LLMs, namely pre-training, adaptation tuning, utilization, and capacity evaluation. Besides, we also summarize the available resources for developing LLMs and discuss the remaining issues for future directions.

Yangkun Chen, Joseph Suarez, Junjie Zhang et al.

We present the results of the second Neural MMO challenge, hosted at IJCAI 2022, which received 1600+ submissions. This competition targets robustness and generalization in multi-agent systems: participants train teams of agents to complete a multi-task objective against opponents not seen during training. The competition combines relatively complex environment design with large numbers of agents in the environment. The top submissions demonstrate strong success on this task using mostly standard reinforcement learning (RL) methods combined with domain-specific engineering. We summarize the competition design and results and suggest that, as an academic community, competitions may be a powerful approach to solving hard problems and establishing a solid benchmark for algorithms. We will open-source our benchmark including the environment wrapper, baselines, a visualization tool, and selected policies for further research.

Junjie Zhang, Yupeng Hou, Ruobing Xie et al.

Recently, there has been an emergence of employing LLM-powered agents as believable human proxies, based on their remarkable decision-making capability. However, existing studies mainly focus on simulating human dialogue. Human non-verbal behaviors, such as item clicking in recommender systems, although implicitly exhibiting user preferences and could enhance the modeling of users, have not been deeply explored. The main reasons lie in the gap between language modeling and behavior modeling, as well as the incomprehension of LLMs about user-item relations. To address this issue, we propose AgentCF for simulating user-item interactions in recommender systems through agent-based collaborative filtering. We creatively consider not only users but also items as agents, and develop a collaborative learning approach that optimizes both kinds of agents together. Specifically, at each time step, we first prompt the user and item agents to interact autonomously. Then, based on the disparities between the agents' decisions and real-world interaction records, user and item agents are prompted to reflect on and adjust the misleading simulations collaboratively, thereby modeling their two-sided relations. The optimized agents can also propagate their preferences to other agents in subsequent interactions, implicitly capturing the collaborative filtering idea. Overall, the optimized agents exhibit diverse interaction behaviors within our framework, including user-item, user-user, item-item, and collective interactions. The results show that these agents can demonstrate personalized behaviors akin to those of real-world individuals, sparking the development of next-generation user behavior simulation.

Xinxin Hu, Haotian Chen, Junjie Zhang et al.

Along with the rapid evolution of mobile communication technologies, such as 5G, there has been a drastically increase in telecom fraud, which significantly dissipates individual fortune and social wealth. In recent years, graph mining techniques are gradually becoming a mainstream solution for detecting telecom fraud. However, the graph imbalance problem, caused by the Pareto principle, brings severe challenges to graph data mining. This is a new and challenging problem, but little previous work has been noticed. In this paper, we propose a Graph ATtention network with COst-sensitive BOosting (GAT-COBO) for the graph imbalance problem. First, we design a GAT-based base classifier to learn the embeddings of all nodes in the graph. Then, we feed the embeddings into a well-designed cost-sensitive learner for imbalanced learning. Next, we update the weights according to the misclassification cost to make the model focus more on the minority class. Finally, we sum the node embeddings obtained by multiple cost-sensitive learners to obtain a comprehensive node representation, which is used for the downstream anomaly detection task. Extensive experiments on two real-world telecom fraud detection datasets demonstrate that our proposed method is effective for the graph imbalance problem, outperforming the state-of-the-art GNNs and GNN-based fraud detectors. In addition, our model is also helpful for solving the widespread over-smoothing problem in GNNs. The GAT-COBO code and datasets are available at https://github.com/xxhu94/GAT-COBO.

Junjie Zhang, Jiafei Lyu, Xiaoteng Ma et al.

Equipped with the trained environmental dynamics, model-based offline reinforcement learning (RL) algorithms can often successfully learn good policies from fixed-sized datasets, even some datasets with poor quality. Unfortunately, however, it can not be guaranteed that the generated samples from the trained dynamics model are reliable (e.g., some synthetic samples may lie outside of the support region of the static dataset). To address this issue, we propose Trajectory Truncation with Uncertainty (TATU), which adaptively truncates the synthetic trajectory if the accumulated uncertainty along the trajectory is too large. We theoretically show the performance bound of TATU to justify its benefits. To empirically show the advantages of TATU, we first combine it with two classical model-based offline RL algorithms, MOPO and COMBO. Furthermore, we integrate TATU with several off-the-shelf model-free offline RL algorithms, e.g., BCQ. Experimental results on the D4RL benchmark show that TATU significantly improves their performance, often by a large margin. Code is available here.

Fusheng Yu, Jiang Li, Xiaoping Wang et al.

Detecting safety clothing and helmets is paramount for ensuring the safety of construction workers. However, the development of deep learning models in this domain has been impeded by the scarcity of high-quality datasets. In this study, we construct a large, complex, and realistic safety clothing and helmet detection (SFCHD) dataset. SFCHD is derived from two authentic chemical plants, comprising 12,373 images, 7 categories, and 50,552 annotations. We partition the SFCHD dataset into training and testing sets with a ratio of 4:1 and validate its utility by applying several classic object detection algorithms. Furthermore, drawing inspiration from spatial and channel attention mechanisms, we design a spatial and channel attention-based low-light enhancement (SCALE) module. SCALE is a plug-and-play component with a high degree of flexibility. Extensive evaluations of the SCALE module on both the ExDark and SFCHD datasets have empirically demonstrated its efficacy in enhancing the performance of detectors under low-light conditions. The dataset and code are publicly available at https://github.com/lijfrank-open/SFCHD-SCALE.

Bochao Liu, Jianghu Lu, Pengju Wang et al.

Deep learning models can achieve high inference accuracy by extracting rich knowledge from massive well-annotated data, but may pose the risk of data privacy leakage in practical deployment. In this paper, we present an effective teacher-student learning approach to train privacy-preserving deep learning models via differentially private data-free distillation. The main idea is generating synthetic data to learn a student that can mimic the ability of a teacher well-trained on private data. In the approach, a generator is first pretrained in a data-free manner by incorporating the teacher as a fixed discriminator. With the generator, massive synthetic data can be generated for model training without exposing data privacy. Then, the synthetic data is fed into the teacher to generate private labels. Towards this end, we propose a label differential privacy algorithm termed selective randomized response to protect the label information. Finally, a student is trained on the synthetic data with the supervision of private labels. In this way, both data privacy and label privacy are well protected in a unified framework, leading to privacy-preserving models. Extensive experiments and analysis clearly demonstrate the effectiveness of our approach.

Yuhao Huang, Sanping Zhou, Junjie Zhang et al.

Efficient representation of point clouds is fundamental for LiDAR-based 3D object detection. While recent grid-based detectors often encode point clouds into either voxels or pillars, the distinctions between these approaches remain underexplored. In this paper, we quantify the differences between the current encoding paradigms and highlight the limited vertical learning within. To tackle these limitations, we introduce a hybrid Voxel-Pillar Fusion network (VPF), which synergistically combines the unique strengths of both voxels and pillars. Specifically, we first develop a sparse voxel-pillar encoder that encodes point clouds into voxel and pillar features through 3D and 2D sparse convolutions respectively, and then introduce the Sparse Fusion Layer (SFL), facilitating bidirectional interaction between sparse voxel and pillar features. Our efficient, fully sparse method can be seamlessly integrated into both dense and sparse detectors. Leveraging this powerful yet straightforward framework, VPF delivers competitive performance, achieving real-time inference speeds on the nuScenes and Waymo Open Dataset. The code will be available.

Junjie Zhang, Zhen Shen, Gang Xiong et al.

Grokking in modular arithmetic has established itself as the quintessential fruit fly experiment, serving as a critical domain for investigating the mechanistic origins of model generalization. Despite its significance, existing research remains narrowly focused on specific local circuits or optimization tuning, largely overlooking the global structural evolution that fundamentally drives this phenomenon. We propose that grokking originates from a spontaneous simplification of internal model structures governed by the principle of parsimony. We integrate causal, spectral, and algorithmic complexity measures alongside Singular Learning Theory to reveal that the transition from memorization to generalization corresponds to the physical collapse of redundant manifolds and deep information compression, offering a novel perspective for understanding the mechanisms of model overfitting and generalization.

Shuang Sun, Huatong Song, Yuhao Wang et al.

Retrieval-augmented generation (RAG) systems have advanced large language models (LLMs) in complex deep search scenarios requiring multi-step reasoning and iterative information retrieval. However, existing approaches face critical limitations that lack high-quality training trajectories or suffer from the distributional mismatches in simulated environments and prohibitive computational costs for real-world deployment. This paper introduces SimpleDeepSearcher, a lightweight yet effective framework that bridges this gap through strategic data engineering rather than complex training paradigms. Our approach synthesizes high-quality training data by simulating realistic user interactions in live web search environments, coupled with a multi-criteria curation strategy that optimizes the diversity and quality of input and output side. Experiments on five benchmarks across diverse domains demonstrate that SFT on only 871 curated samples yields significant improvements over RL-based baselines. Our work establishes SFT as a viable pathway by systematically addressing the data-scarce bottleneck, offering practical insights for efficient deep search systems. Our code is available at https://github.com/RUCAIBox/SimpleDeepSearcher.

Junjie Zhang, Feng Zhao, Hanqiang Liu et al.

The booming remote sensing (RS) technology is giving rise to a novel multimodality generalization task, which requires the model to overcome data heterogeneity while possessing powerful cross-scene generalization ability. Moreover, most vision-language models (VLMs) usually describe surface materials in RS images using universal texts, lacking proprietary linguistic prior knowledge specific to different RS vision modalities. In this work, we formalize RS multimodality generalization (RSMG) as a learning paradigm, and propose a frequency-aware vision-language multimodality generalization network (FVMGN) for RS image classification. Specifically, a diffusion-based training-test-time augmentation (DTAug) strategy is designed to reconstruct multimodal land-cover distributions, enriching input information for FVMGN. Following that, to overcome multimodal heterogeneity, a multimodal wavelet disentanglement (MWDis) module is developed to learn cross-domain invariant features by resampling low and high frequency components in the frequency domain. Considering the characteristics of RS vision modalities, shared and proprietary class texts is designed as linguistic inputs for the transformer-based text encoder to extract diverse text features. For multimodal vision inputs, a spatial-frequency-aware image encoder (SFIE) is constructed to realize local-global feature reconstruction and representation. Finally, a multiscale spatial-frequency feature alignment (MSFFA) module is suggested to construct a unified semantic space, ensuring refined multiscale alignment of different text and vision features in spatial and frequency domains. Extensive experiments show that FVMGN has the excellent multimodality generalization ability compared with state-of-the-art (SOTA) methods.

Shunyu Liu, Wenkai Fang, Zetian Hu et al.

Large Language Models (LLMs) have demonstrated unprecedented generative capabilities, yet their alignment with human values remains critical for ensuring helpful and harmless deployments. While Reinforcement Learning from Human Feedback (RLHF) has emerged as a powerful paradigm for aligning LLMs with human preferences, its reliance on complex reward modeling introduces inherent trade-offs in computational efficiency and training stability. In this context, Direct Preference Optimization (DPO) has recently gained prominence as a streamlined alternative that directly optimizes LLMs using human preferences, thereby circumventing the need for explicit reward modeling. Owing to its theoretical elegance and computational efficiency, DPO has rapidly attracted substantial research efforts exploring its various implementations and applications. However, this field currently lacks systematic organization and comparative analysis. In this survey, we conduct a comprehensive overview of DPO and introduce a novel taxonomy, categorizing previous works into four key dimensions: data strategy, learning framework, constraint mechanism, and model property. We further present a rigorous empirical analysis of DPO variants across standardized benchmarks. Additionally, we discuss real-world applications, open challenges, and future directions for DPO. This work delivers both a conceptual framework for understanding DPO and practical guidance for practitioners, aiming to advance robust and generalizable alignment paradigms. All collected resources are available and will be continuously updated at https://github.com/liushunyu/awesome-direct-preference-optimization.

Yaoqin Ye, Junjie Zhang, Hongwei Shi

The task of medical image recognition is notably complicated by the presence of varied and multiple pathological indications, presenting a unique challenge in multi-label classification with unseen labels. This complexity underlines the need for computer-aided diagnosis methods employing multi-label zero-shot learning. Recent advancements in pre-trained vision-language models (VLMs) have showcased notable zero-shot classification abilities on medical images. However, these methods have limitations on leveraging extensive pre-trained knowledge from broader image datasets, and often depend on manual prompt construction by expert radiologists. By automating the process of prompt tuning, prompt learning techniques have emerged as an efficient way to adapt VLMs to downstream tasks. Yet, existing CoOp-based strategies fall short in performing class-specific prompts on unseen categories, limiting generalizability in fine-grained scenarios. To overcome these constraints, we introduce a novel prompt generation approach inspirited by text generation in natural language processing (NLP). Our method, named Pseudo-Prompt Generating (PsPG), capitalizes on the priori knowledge of multi-modal features. Featuring a RNN-based decoder, PsPG autoregressively generates class-tailored embedding vectors, i.e., pseudo-prompts. Comparative evaluations on various multi-label chest radiograph datasets affirm the superiority of our approach against leading medical vision-language and multi-label prompt learning methods. The source code is available at https://github.com/fallingnight/PsPG

Junjie Zhang, Guozheng Ma, Shunyu Liu et al.

Recent advances in Reinforcement Learning (RL) have underscored its potential for incentivizing reasoning capabilities of Large Language Models (LLMs). However, existing step-level efforts suffer from costly annotations that limit domain coverage, while scalar scores further impose an information bottleneck, offering insufficient semantic bandwidth to improve intermediate decisions. Alternative language-critique approaches, which rely on frozen or external critics, provide richer textual feedback but lack the scalability needed for sustained policy improvement. In this work, we propose language-driven stepwise trajectory redirection, termed as STRIDE, a novel training framework that shifts process supervision from scalar rewards to learnable stepwise language feedback. Specifically, we co-train a generator and a generative verifier using only outcome-based rewards, eliminating external annotations, while delivering sustained policy improvement through jointly aligned verifier training. The verifier's stepwise language critiques explicitly localize and explain failures, enabling the generator to redirect reasoning trajectories at intermediate steps toward alternative decisions. The trajectory redirection design guarantees harmless policy improvement, even under noisy or suboptimal verifier feedback. Experiments on diverse reasoning benchmarks show that STRIDE significantly outperforms state-of-the-art baselines, as well as achieving breakthroughs on zero-pass-rate problems where scalar methods yield no learning signal in our ablation studies, demonstrating the effectiveness of learnable stepwise language feedback for enhancing LLM reasoning.

Junjie Zhang, Jingyi Xi, Zhuoyang Song et al.

Training large language models (LLMs) to act as autonomous agents for multi-turn, long-horizon tasks remains significant challenges in scalability and training efficiency. To address this, we introduce L-Zero (L0), a scalable, end-to-end training pipeline for general-purpose agents. Featuring a low-cost, extensible, and sandboxed concurrent agent worker pool, L0 lowers the barrier for applying reinforcement learning in complex environments. We also introduce NB-Agent, the agent scaffold within L0, which operates in a "code-as-action" fashion via a Read-Eval-Print-Loop (REPL). We evaluate L0 on factuality question-answering benchmarks. Our experiments demonstrate that a base model can develop robust problem-solving skills using solely Reinforcement Learning with Verifiable Rewards (RLVR). On the Qwen2.5-7B-Instruct model, our method boosts accuracy on SimpleQA from 30 % to 80 % and on HotpotQA from 22 % to 41 %. We have open-sourced the entire L0 system, including our L0 series models, the NB-Agent, a complete training pipeline, and the corresponding training recipes on (https://github.com/cmriat/l0).

Junjie Zhang, Rushuai Yang, Shunyu Liu et al.

In this work, we establish a novel theoretical connection between supervised fine-tuning and offline reinforcement learning under the token-level Markov decision process, revealing that large language models indeed learn an implicit $Q$-function for inference. Through this theoretical lens, we demonstrate that the widely used beam search method suffers from unacceptable over-optimism, where inference errors are inevitably amplified due to inflated $Q$-value estimations of suboptimal steps. To address this limitation, we propose Supervised Optimism Correction(SOC), which introduces a simple yet effective auxiliary loss for token-level $Q$-value estimations during supervised fine-tuning. Specifically, the auxiliary loss employs implicit value regularization to boost model confidence in expert-demonstrated responses, thereby suppressing over-optimism toward insufficiently supervised responses. Extensive experiments on mathematical reasoning benchmarks, including GSM8K, MATH, and GAOKAO, showcase the superiority of the proposed SOC with beam search across a series of open-source models.

Ting Sun, Junjie Zhang, Xiao Yan et al.

Modern Large Foundation Model (LFM) training has transformed the data pipeline from a static ingestion layer into a dynamic component that must co-evolve with the training process. Existing systems are ill-equipped: colocated dataloaders offer no failure isolation, while message queue-based disaggregated dataloaders operate on a record/offset abstraction that cannot express the batch-level semantics required by distributed training. We present Lakestream, a brokerless, object-store-native training data plane with three key properties. First, it introduces the Transactional Global Batch (TGB), which builds on lakehouse-style ACID storage semantics and extends them with training-specific consistency, including atomic all-rank batch visibility, a globally ordered step sequence, checkpoint-aligned lifecycle management, and end-to-end exactly-once recovery. Second, it realizes recovery and retention directly in the storage layer, by inlining producer state in the manifest and tying reclamation to distributed checkpoint state. Third, its Decentralized Adaptive Commit (DAC) algorithm sustains stable ingestion throughput as the manifest grows, without any inter-producer communication. Evaluations on large-scale multimodal pre-training and SFT workloads using 64 GPUs show that Lakestream outperforms colocated dataloader throughput while providing full failure isolation, outperforms Apache Kafka in ingestion throughput, and achieves lower consumer read latency than Kafka.

Yupeng Hou, Junjie Zhang, Zihan Lin et al.

Recently, large language models (LLMs) (e.g., GPT-4) have demonstrated impressive general-purpose task-solving abilities, including the potential to approach recommendation tasks. Along this line of research, this work aims to investigate the capacity of LLMs that act as the ranking model for recommender systems. We first formalize the recommendation problem as a conditional ranking task, considering sequential interaction histories as conditions and the items retrieved by other candidate generation models as candidates. To solve the ranking task by LLMs, we carefully design the prompting template and conduct extensive experiments on two widely-used datasets. We show that LLMs have promising zero-shot ranking abilities but (1) struggle to perceive the order of historical interactions, and (2) can be biased by popularity or item positions in the prompts. We demonstrate that these issues can be alleviated using specially designed prompting and bootstrapping strategies. Equipped with these insights, zero-shot LLMs can even challenge conventional recommendation models when ranking candidates are retrieved by multiple candidate generators. The code and processed datasets are available at https://github.com/RUCAIBox/LLMRank.

Junjie Zhang, Ruobing Xie, Yupeng Hou et al.

In the past decades, recommender systems have attracted much attention in both research and industry communities, and a large number of studies have been devoted to developing effective recommendation models. Basically speaking, these models mainly learn the underlying user preference from historical behavior data, and then estimate the user-item matching relationships for recommendations. Inspired by the recent progress on large language models (LLMs), we take a different approach to developing the recommendation models, considering recommendation as instruction following by LLMs. The key idea is that the preferences or needs of a user can be expressed in natural language descriptions (called instructions), so that LLMs can understand and further execute the instruction for fulfilling the recommendation task. Instead of using public APIs of LLMs, we instruction tune an open-source LLM (3B Flan-T5-XL), in order to better adapt LLMs to recommender systems. For this purpose, we first design a general instruction format for describing the preference, intention, task form and context of a user in natural language. Then we manually design 39 instruction templates and automatically generate a large amount of user-personalized instruction data (252K instructions) with varying types of preferences and intentions. To demonstrate the effectiveness of our approach, we instantiate the instruction templates into several widely-studied recommendation (or search) tasks, and conduct extensive experiments on these tasks with real-world datasets. Experiment results show that the proposed approach can outperform several competitive baselines, including the powerful GPT-3.5, on these evaluation tasks. Our approach sheds light on developing more user-friendly recommender systems, in which users can freely communicate with the system and obtain more accurate recommendations via natural language instructions.

Junjie Zhang, Tianci Hu, Xiaoshui Huang et al.

Evaluating the performance of Multi-modal Large Language Models (MLLMs), integrating both point cloud and language, presents significant challenges. The lack of a comprehensive assessment hampers determining whether these models truly represent advancements, thereby impeding further progress in the field. Current evaluations heavily rely on classification and caption tasks, falling short in providing a thorough assessment of MLLMs. A pressing need exists for a more sophisticated evaluation method capable of thoroughly analyzing the spatial understanding and expressive capabilities of these models. To address these issues, we introduce a scalable 3D benchmark, accompanied by a large-scale instruction-tuning dataset known as 3DBench, providing an extensible platform for a comprehensive evaluation of MLLMs. Specifically, we establish the benchmark that spans a wide range of spatial and semantic scales, from object-level to scene-level, addressing both perception and planning tasks. Furthermore, we present a rigorous pipeline for automatically constructing scalable 3D instruction-tuning datasets, covering 10 diverse multi-modal tasks with more than 0.23 million QA pairs generated in total. Thorough experiments evaluating trending MLLMs, comparisons against existing datasets, and variations of training protocols demonstrate the superiority of 3DBench, offering valuable insights into current limitations and potential research directions.

Junjie Zhang, Beichen Zhang, Wenqi Sun et al.

To develop effective sequential recommender systems, numerous methods have been proposed to model historical user behaviors. Despite the effectiveness, these methods share the same fast thinking paradigm. That is, for making recommendations, these methods typically encodes user historical interactions to obtain user representations and directly match these representations with candidate item representations. However, due to the limited capacity of traditional lightweight recommendation models, this one-step inference paradigm often leads to suboptimal performance. To tackle this issue, we present a novel slow thinking recommendation model, named STREAM-Rec. Our approach is capable of analyzing historical user behavior, generating a multi-step, deliberative reasoning process, and ultimately delivering personalized recommendations. In particular, we focus on two key challenges: (1) identifying the suitable reasoning patterns in recommender systems, and (2) exploring how to effectively stimulate the reasoning capabilities of traditional recommenders. To this end, we introduce a three-stage training framework. In the first stage, the model is pretrained on large-scale user behavior data to learn behavior patterns and capture long-range dependencies. In the second stage, we design an iterative inference algorithm to annotate suitable reasoning traces by progressively refining the model predictions. This annotated data is then used to fine-tune the model. Finally, in the third stage, we apply reinforcement learning to further enhance the model generalization ability. Extensive experiments validate the effectiveness of our proposed method.

Ruiyang Ha, Songyi Jiang, Bin Li et al.

Conventional person re-identification (ReID) research is often limited to single-modality sensor data from static cameras, which fails to address the complexities of real-world scenarios where multi-modal signals are increasingly prevalent. For instance, consider an urban ReID system integrating stationary RGB cameras, nighttime infrared sensors, and UAVs equipped with dynamic tracking capabilities. Such systems face significant challenges due to variations in camera perspectives, lighting conditions, and sensor modalities, hindering effective person ReID. To address these challenges, we introduce the MP-ReID benchmark, a novel dataset designed specifically for multi-modality and multi-platform ReID. This benchmark uniquely compiles data from 1,930 identities across diverse modalities, including RGB, infrared, and thermal imaging, captured by both UAVs and ground-based cameras in indoor and outdoor environments. Building on this benchmark, we introduce Uni-Prompt ReID, a framework with specific-designed prompts, tailored for cross-modality and cross-platform scenarios. Our method consistently outperforms state-of-the-art approaches, establishing a robust foundation for future research in complex and dynamic ReID environments. Our dataset are available at:https://mp-reid.github.io/.

Junjie Zhang, Guozheng Ma, Shunyu Liu et al.

Reinforcement Learning with Verifiable Rewards (RLVR) has emerged as a powerful learn-to-reason paradigm for Large Reasoning Models to tackle complex tasks. However, current RLVR paradigm is still not efficient enough, as it works in a trial-and-error manner. To perform better, the model needs to explore the reward space by numerously generating responses and learn from fragmented reward signals, blind to the overall reward patterns. Fortunately, verifiable rewards make the natural language description of the reward function possible, and meanwhile, LLMs have demonstrated strong in-context learning ability. This motivates us to explore if Large Reasoning Models can benefit from a motivation of the task, i.e., awareness of the reward function, during the reinforcement finetuning process, as we humans sometimes do when learning. In this paper, we introduce Motivation-enhanced Reinforcement Finetuning (MeRF), an intuitive yet effective method enhancing reinforcement finetuning of LLMs by involving ``telling LLMs rules of the game''. Specifically, MeRF directly injects the reward specification into the prompt, which serves as an in-context motivation for the model to be aware of the optimization objective. This simple modification leverages the in-context learning ability of LLMs, aligning generation with optimization, thereby incentivizing the model to generate desired outputs from both inner motivation and external reward. Empirical evaluations demonstrate that MeRF achieves substantial performance gains over RLVR baseline. Moreover, ablation studies show that MeRF performs better with greater consistency between the in-context motivation and the external reward function, while the model also demonstrates an ability to adapt to misleading motivations through reinforcement finetuning.

Junjie Zhang, Zheming Zhang, Huachen Xiang et al.

Physical function monitoring (PFM) plays a crucial role in healthcare especially for the elderly. Traditional assessment methods such as the Short Physical Performance Battery (SPPB) have failed to capture the full dynamic characteristics of physical function. Wearable sensors such as smart wristbands offer a promising solution to this issue. However, challenges exist, such as the computational complexity of machine learning methods and inadequate information capture. This paper proposes a multi-modal PFM framework based on an improved TimeMAE, which compresses time-series data into a low-dimensional latent space and integrates a self-enhanced attention module. This framework achieves effective monitoring of physical health, providing a solution for real-time and personalized assessment. The method is validated using the NHATS dataset, and the results demonstrate an accuracy of 70.6% and an AUC of 82.20%, surpassing other state-of-the-art time-series classification models.

Zezhou Zhang, Songxin Zhang, Xiao Xiong et al.

VLA architectures that pair a pretrained VLM with a flow-matching action expert have emerged as a strong paradigm for language-conditioned manipulation. Yet the VLM, optimized for semantic abstraction and typically conditioned on static visual observations, tends to attenuate fine-grained geometric cues and often lacks explicit temporal evidence for the action expert. Prior work mitigates this by injecting auxiliary visual features, but existing approaches either focus on static spatial representations or require substantial architectural modifications to accommodate temporal inputs, leaving temporal information underexplored. We propose Plug-in Visual Injection (PVI), a lightweight, encoder-agnostic module that attaches to a pretrained action expert and injects auxiliary visual representations via zero-initialized residual pathways, preserving pretrained behavior with only single-stage fine-tuning. Using PVI, we obtain consistent gains over the base policy and a range of competitive alternative injection strategies, and our controlled study shows that temporal video features (V-JEPA2) outperform strong static image features (DINOv2), with the largest gains on multi-phase tasks requiring state tracking and coordination. Real-robot experiments on long-horizon bimanual cloth folding further demonstrate the practicality of PVI beyond simulation.

Junjie Zhang, Zhen Shen, Gang Xiong et al.

The evolution of intelligence in artificial systems provides a unique opportunity to identify universal computational principles. Here we show that large language models spontaneously develop synergistic cores where information integration exceeds individual parts remarkably similar to the human brain. Using Integrated Information Decomposition across multiple architectures we find that middle layers exhibit synergistic processing while early and late layers rely on redundancy. This organization is dynamic and emerges as a physical phase transition as task difficulty increases. Crucially ablating synergistic components causes catastrophic performance loss confirming their role as the physical entity of abstract reasoning and bridging artificial and biological intelligence.

Haisheng Su, Junjie Zhang, Feixiang Song et al.

Detecting 3D objects accurately from multi-view 2D images is a challenging yet essential task in the field of autonomous driving. Current methods resort to integrating depth prediction to recover the spatial information for object query decoding, which necessitates explicit supervision from LiDAR points during the training phase. However, the predicted depth quality is still unsatisfactory such as depth discontinuity of object boundaries and indistinction of small objects, which are mainly caused by the sparse supervision of projected points and the use of high-level image features for depth prediction. Besides, cross-view consistency and scale invariance are also overlooked in previous methods. In this paper, we introduce Frequency-aware Positional Depth Embedding (FreqPDE) to equip 2D image features with spatial information for 3D detection transformer decoder, which can be obtained through three main modules. Specifically, the Frequency-aware Spatial Pyramid Encoder (FSPE) constructs a feature pyramid by combining high-frequency edge clues and low-frequency semantics from different levels respectively. Then the Cross-view Scale-invariant Depth Predictor (CSDP) estimates the pixel-level depth distribution with cross-view and efficient channel attention mechanism. Finally, the Positional Depth Encoder (PDE) combines the 2D image features and 3D position embeddings to generate the 3D depth-aware features for query decoding. Additionally, hybrid depth supervision is adopted for complementary depth learning from both metric and distribution aspects. Extensive experiments conducted on the nuScenes dataset demonstrate the effectiveness and superiority of our proposed method.

Chenghao Wu, Ruiyang Ren, Junjie Zhang et al.

While modern recommender systems are instrumental in navigating information abundance, they remain fundamentally limited by static user modeling and reactive decision-making paradigms. Current large language model (LLM)-based agents inherit these shortcomings through their overreliance on heuristic pattern matching, yielding recommendations prone to shallow correlation bias, limited causal inference, and brittleness in sparse-data scenarios. We introduce STARec, a slow-thinking augmented agent framework that endows recommender systems with autonomous deliberative reasoning capabilities. Each user is modeled as an agent with parallel cognitions: fast response for immediate interactions and slow reasoning that performs chain-of-thought rationales. To cultivate intrinsic slow thinking, we develop anchored reinforcement training - a two-stage paradigm combining structured knowledge distillation from advanced reasoning models with preference-aligned reward shaping. This hybrid approach scaffolds agents in acquiring foundational capabilities (preference summarization, rationale generation) while enabling dynamic policy adaptation through simulated feedback loops. Experiments on MovieLens 1M and Amazon CDs benchmarks demonstrate that STARec achieves substantial performance gains compared with state-of-the-art baselines, despite using only 0.4% of the full training data.

Xi Chen, Jianchuan Yang, Junjie Zhang et al.

Physics-Informed Neural Networks offer a powerful framework for solving PDEs by embedding physical laws into the learning process. However, when applied to domains with irregular boundaries, PINNs often suffer from instability and slow convergence, which stems from (1) inconsistent normalization due to geometric anisotropy, (2) inaccurate boundary enforcements, and (3) imbalanced loss term competition. A common workaround is to map the domain to a regular space. Yet, conventional mapping methods rely on case-specific meshes, define Jacobians at pre-specified fixed nodes, reformulate PDEs via the chain rule-making them incompatible with modern automatic differentiation, tensor-based frameworks. To bridge this gap, we propose JacobiNet, a learning-based coordinate-transformed PINN framework that unifies domain mapping and PDE solving within an end-to-end differentiable architecture. Leveraging lightweight MLPs, JacobiNet learns continuous, differentiable mappings, enables direct Jacobian computation via autograd, shares computation graph with downstream PINNs. Its continuous nature and built-in Jacobian eliminate the need for meshing, explicit Jacobians computation/ storage, and PDE reformulation, while unlocking geometric-editing operations, reducing the mapping cost. Separating physical modeling from geometric complexity, JacobiNet (1) addresses normalization challenges in the original anisotropic coordinates, (2) facilitates hard constraints of boundary conditions, and (3) mitigates the long-standing imbalance among loss terms. Evaluated on various PDEs, JacobiNet reduces the L2 error from 0.11-0.73 to 0.01-0.09. In vessel-like domains with varying shapes, JacobiNet enables millisecond-level mapping inference for unseen geometries, improves prediction accuracy by an average of 3.65*, while delivering over 10* speed up-demonstrating strong generalization, accuracy, and efficiency.

Junjie Zhang, Zhimin Zong, Lin Gu et al.

The evolution of colour vision is captivating, as it reveals the adaptive strategies of extinct species while simultaneously inspiring innovations in modern imaging technology. In this study, we present a simplified model of visual transduction in the retina, introducing a novel opsin layer. We quantify evolutionary pressures by measuring machine vision recognition accuracy on colour images shaped by specific opsins. Building on this, we develop an evolutionary conservation optimisation algorithm to reconstruct the spectral sensitivity of opsins, enabling mutation-driven adaptations to to more effectively spot fruits or predators. This model condenses millions of years of evolution within seconds on GPU, providing an experimental framework to test long-standing hypotheses in evolutionary biology , such as vision of early mammals, primate trichromacy from gene duplication, retention of colour blindness, blue-shift of fish rod and multiple rod opsins with bioluminescence. Moreover, the model enables speculative explorations of hypothetical species, such as organisms with eyes adapted to the conditions on Mars. Our findings suggest a minimalist yet effective approach to task-specific camera filter design, optimising the spectral response function to meet application-driven demands. The code will be made publicly available upon acceptance.

Shilu Yuan, Dongfeng Li, Wei Liu et al.

Fine-grained urban flow inference (FUFI) is a crucial transportation service aimed at improving traffic efficiency and safety. FUFI can infer fine-grained urban traffic flows based solely on observed coarse-grained data. However, most of existing methods focus on the influence of single-scale static geographic information on FUFI, neglecting the interactions and dynamic information between different-scale regions within the city. Different-scale geographical features can capture redundant information from the same spatial areas. In order to effectively learn multi-scale information across time and space, we propose an effective fine-grained urban flow inference model called UrbanMSR, which uses self-supervised contrastive learning to obtain dynamic multi-scale representations of neighborhood-level and city-level geographic information, and fuses multi-scale representations to improve fine-grained accuracy. The fusion of multi-scale representations enhances fine-grained. We validate the performance through extensive experiments on three real-world datasets. The resutls compared with state-of-the-art methods demonstrate the superiority of the proposed model.

Huaxi Huang, Junjie Zhang, Jian Zhang et al.

The predicament in semi-supervised few-shot learning (SSFSL) is to maximize the value of the extra unlabeled data to boost the few-shot learner. In this paper, we propose a Poisson Transfer Network (PTN) to mine the unlabeled information for SSFSL from two aspects. First, the Poisson Merriman Bence Osher (MBO) model builds a bridge for the communications between labeled and unlabeled examples. This model serves as a more stable and informative classifier than traditional graph-based SSFSL methods in the message-passing process of the labels. Second, the extra unlabeled samples are employed to transfer the knowledge from base classes to novel classes through contrastive learning. Specifically, we force the augmented positive pairs close while push the negative ones distant. Our contrastive transfer scheme implicitly learns the novel-class embeddings to alleviate the over-fitting problem on the few labeled data. Thus, we can mitigate the degeneration of embedding generality in novel classes. Extensive experiments indicate that PTN outperforms the state-of-the-art few-shot and SSFSL models on miniImageNet and tieredImageNet benchmark datasets.

Junjie Zhang, Cong Zhang, Neal N. Xiong

Inverse Distance Weighted models (IDW) have been widely used for predicting and modeling multidimensional space in multimodal industrial processes. However, the more complex the structure of multidimensional space, the lower the performance of IDW models, and real industrial datasets tend to have more complex spatial structure. To solve this problem, a new framework for spatial prediction and modeling based on deep reinforcement learning network is proposed. In the proposed framework, the internal relationship between state and action is enhanced by reusing the state values in the Q network, and the convergence rate and stability of the deep reinforcement learning network are improved. The improved deep reinforcement learning network is then used to search for and learn the hyperparameters of each sample point in the inverse distance weighted model. These hyperparameters can reflect the spatial structure of the current industrial dataset to some extent. Then a spatial distribution of hyperparameters is constructed based on the learned hyperparameters. Each interpolation point obtains corresponding hyperparameters from the hyperparametric spatial distribution and brings them into the classical IDW models for prediction, thus achieving differential spatial prediction and modeling. The simulation results show that the proposed framework is suitable for real industrial datasets with complex spatial structure characteristics and is more accurate than current IDW models in spatial prediction.

Huaxi Huang, Junjie Zhang, Jian Zhang et al.

The challenges of high intra-class variance yet low inter-class fluctuations in fine-grained visual categorization are more severe with few labeled samples, \textit{i.e.,} Fine-Grained categorization problems under the Few-Shot setting (FGFS). High-order features are usually developed to uncover subtle differences between sub-categories in FGFS, but they are less effective in handling the high intra-class variance. In this paper, we propose a Target-Oriented Alignment Network (TOAN) to investigate the fine-grained relation between the target query image and support classes. The feature of each support image is transformed to match the query ones in the embedding feature space, which reduces the disparity explicitly within each category. Moreover, different from existing FGFS approaches devise the high-order features over the global image with less explicit consideration of discriminative parts, we generate discriminative fine-grained features by integrating compositional concept representations to global second-order pooling. Extensive experiments are conducted on four fine-grained benchmarks to demonstrate the effectiveness of TOAN compared with the state-of-the-art models.

Junjie Zhang, Minghao Ye, Zehua Guo et al.

Traditional Traffic Engineering (TE) solutions can achieve the optimal or near-optimal performance by rerouting as many flows as possible. However, they do not usually consider the negative impact, such as packet out of order, when frequently rerouting flows in the network. To mitigate the impact of network disturbance, one promising TE solution is forwarding the majority of traffic flows using Equal-Cost Multi-Path (ECMP) and selectively rerouting a few critical flows using Software-Defined Networking (SDN) to balance link utilization of the network. However, critical flow rerouting is not trivial because the solution space for critical flow selection is enormous. Moreover, it is impossible to design a heuristic algorithm for this problem based on fixed and simple rules, since rule-based heuristics are unable to adapt to the changes of the traffic matrix and network dynamics. In this paper, we propose CFR-RL (Critical Flow Rerouting-Reinforcement Learning), a Reinforcement Learning-based scheme that learns a policy to select critical flows for each given traffic matrix automatically. CFR-RL then reroutes these selected critical flows to balance link utilization of the network by formulating and solving a simple Linear Programming (LP) problem. Extensive evaluations show that CFR-RL achieves near-optimal performance by rerouting only 10%-21.3% of total traffic.

Junjie Zhang, Lingqiao Liu, Peng Wang et al.

Real-world visual data often exhibits a long-tailed distribution, where some ''head'' classes have a large number of samples, yet only a few samples are available for ''tail'' classes. Such imbalanced distribution causes a great challenge for learning a deep neural network, which can be boiled down into a dilemma: on the one hand, we prefer to increase the exposure of tail class samples to avoid the excessive dominance of head classes in the classifier training. On the other hand, oversampling tail classes makes the network prone to over-fitting, since head class samples are often consequently under-represented. To resolve this dilemma, in this paper, we propose a simple-yet-effective auxiliary learning approach. The key idea is to split a network into a classifier part and a feature extractor part, and then employ different training strategies for each part. Specifically, to promote the awareness of tail-classes, a class-balanced sampling scheme is utilised for training both the classifier and the feature extractor. For the feature extractor, we also introduce an auxiliary training task, which is to train a classifier under the regular random sampling scheme. In this way, the feature extractor is jointly trained from both sampling strategies and thus can take advantage of all training data and avoid the over-fitting issue. Apart from this basic auxiliary task, we further explore the benefit of using self-supervised learning as the auxiliary task. Without using any bells and whistles, our model achieves superior performance over the state-of-the-art solutions.

Huaxi Huang, Junjie Zhang, Jian Zhang et al.

Deep neural networks have demonstrated advanced abilities on various visual classification tasks, which heavily rely on the large-scale training samples with annotated ground-truth. However, it is unrealistic always to require such annotation in real-world applications. Recently, Few-Shot learning (FS), as an attempt to address the shortage of training samples, has made significant progress in generic classification tasks. Nonetheless, it is still challenging for current FS models to distinguish the subtle differences between fine-grained categories given limited training data. To filling the classification gap, in this paper, we address the Few-Shot Fine-Grained (FSFG) classification problem, which focuses on tackling the fine-grained classification under the challenging few-shot learning setting. A novel low-rank pairwise bilinear pooling operation is proposed to capture the nuanced differences between the support and query images for learning an effective distance metric. Moreover, a feature alignment layer is designed to match the support image features with query ones before the comparison. We name the proposed model Low-Rank Pairwise Alignment Bilinear Network (LRPABN), which is trained in an end-to-end fashion. Comprehensive experimental results on four widely used fine-grained classification datasets demonstrate that our LRPABN model achieves the superior performances compared to state-of-the-art methods.

Huaxi Huang, Junjie Zhang, Jian Zhang et al.

The recognition ability of human beings is developed in a progressive way. Usually, children learn to discriminate various objects from coarse to fine-grained with limited supervision. Inspired by this learning process, we propose a simple yet effective model for the Few-Shot Fine-Grained (FSFG) recognition, which tries to tackle the challenging fine-grained recognition task using meta-learning. The proposed method, named Pairwise Alignment Bilinear Network (PABN), is an end-to-end deep neural network. Unlike traditional deep bilinear networks for fine-grained classification, which adopt the self-bilinear pooling to capture the subtle features of images, the proposed model uses a novel pairwise bilinear pooling to compare the nuanced differences between base images and query images for learning a deep distance metric. In order to match base image features with query image features, we design feature alignment losses before the proposed pairwise bilinear pooling. Experiment results on four fine-grained classification datasets and one generic few-shot dataset demonstrate that the proposed model outperforms both the state-ofthe-art few-shot fine-grained and general few-shot methods.

Junjie Zhang, Yong Xia, Yanning Zhang

Detection of pulmonary nodules on chest CT is an essential step in the early diagnosis of lung cancer, which is critical for best patient care. Although a number of computer-aided nodule detection methods have been published in the literature, these methods still have two major drawbacks: missing out true nodules during the detection of nodule candidates and less-accurate identification of nodules from non-nodule. In this paper, we propose an automated pulmonary nodule detection algorithm that jointly combines progressive resolution and hierarchical saliency. Specifically, we design a 3D progressive resolution-based densely dilated FCN, namely the progressive resolution network (PRN), to detect nodule candidates inside the lung, and construct a densely dilated 3D CNN with hierarchical saliency, namely the hierarchical saliency network (HSN), to simultaneously identify genuine nodules from those candidates and estimate the diameters of nodules. We evaluated our algorithm on the benchmark LUng Nodule Analysis 2016 (LUNA16) dataset and achieved a state-of-the-art detection score. Our results suggest that the proposed algorithm can effectively detect pulmonary nodules on chest CT and accurately estimate their diameters.

Junjie Zhang, Qi Wu, Chunhua Shen et al.

Despite significant progress in a variety of vision-and-language problems, developing a method capable of asking intelligent, goal-oriented questions about images is proven to be an inscrutable challenge. Towards this end, we propose a Deep Reinforcement Learning framework based on three new intermediate rewards, namely goal-achieved, progressive and informativeness that encourage the generation of succinct questions, which in turn uncover valuable information towards the overall goal. By directly optimizing for questions that work quickly towards fulfilling the overall goal, we avoid the tendency of existing methods to generate long series of insane queries that add little value. We evaluate our model on the GuessWhat?! dataset and show that the resulting questions can help a standard Guesser identify a specific object in an image at a much higher success rate.

Junjie Zhang, Qi Wu, Jian Zhang et al.

The number of social images has exploded by the wide adoption of social networks, and people like to share their comments about them. These comments can be a description of the image, or some objects, attributes, scenes in it, which are normally used as the user-provided tags. However, it is well-known that user-provided tags are incomplete and imprecise to some extent. Directly using them can damage the performance of related applications, such as the image annotation and retrieval. In this paper, we propose to learn an image annotation model and refine the user-provided tags simultaneously in a weakly-supervised manner. The deep neural network is utilized as the image feature learning and backbone annotation model, while visual consistency, semantic dependency, and user-error sparsity are introduced as the constraints at the batch level to alleviate the tag noise. Therefore, our model is highly flexible and stable to handle large-scale image sets. Experimental results on two benchmark datasets indicate that our proposed model achieves the best performance compared to the state-of-the-art methods.

Haoyu Xu, Chongyang Gu, Han Zhou et al.

The absence of an appropriate text classification corpus makes the massive amount of online job information unusable for labor market analysis. This paper presents JCTC, a large job posting corpus for text classification. In JCTC construction framework, a formal specification issued by the Chinese central government is chosen as the classification standard. The unsupervised learning (WE-cos), supervised learning algorithm (SVM) and human judgements are all used in the construction process. JCTC has 102581 online job postings distributed in 465 categories. The method proposed here can not only ameliorate the high demands on people's skill and knowledge, but reduce the subjective influences as well. Besides, the method is not limited in Chinese. We benchmark five state-of-the-art deep learning approaches on JCTC providing baseline results for future studies. JCTC might be the first job posting corpus for text classification and the largest one in Chinese. With the help of JCTC, related organizations are able to monitor, analyze and predict the labor market in a comprehensive, accurate and timely manner.

Junjie Zhang, Qi Wu, Chunhua Shen et al.

Deep convolution neural networks (CNN) have demonstrated advanced performance on single-label image classification, and various progress also have been made to apply CNN methods on multi-label image classification, which requires to annotate objects, attributes, scene categories etc. in a single shot. Recent state-of-the-art approaches to multi-label image classification exploit the label dependencies in an image, at global level, largely improving the labeling capacity. However, predicting small objects and visual concepts is still challenging due to the limited discrimination of the global visual features. In this paper, we propose a Regional Latent Semantic Dependencies model (RLSD) to address this problem. The utilized model includes a fully convolutional localization architecture to localize the regions that may contain multiple highly-dependent labels. The localized regions are further sent to the recurrent neural networks (RNN) to characterize the latent semantic dependencies at the regional level. Experimental results on several benchmark datasets show that our proposed model achieves the best performance compared to the state-of-the-art models, especially for predicting small objects occurred in the images. In addition, we set up an upper bound model (RLSD+ft-RPN) using bounding box coordinates during training, the experimental results also show that our RLSD can approach the upper bound without using the bounding-box annotations, which is more realistic in the real world.