Xingxing Wang

Zhen Zhang, Luowen Liu, Wanzhi Zhang et al.

With the advancement of machine learning, an increasing number of studies are employing automated mechanism design (AMD) methods for optimal auction design. However, all previous AMD architectures designed to generate optimal mechanisms that satisfy near dominant strategy incentive compatibility (DSIC) fail to achieve deterministic allocation, and some also lack anonymity, thereby impacting the efficiency and fairness of advertising allocation. This has resulted in a notable discrepancy between the previous AMD architectures for generating near-DSIC optimal mechanisms and the demands of real-world advertising scenarios. In this paper, we prove that in all online advertising scenarios, previous non-deterministic allocation methods lead to the non-existence of feasible solutions, resulting in a gap between the rounded solution and the optimal solution. Furthermore, we propose JTransNet, a transformer-based neural network architecture, designed for optimal deterministic-allocation and anonymous joint auction design. Although the deterministic allocation module in JTransNet is designed for the latest joint auction scenarios, it can be applied to other non-deterministic AMD architectures with minor modifications. Additionally, our offline and online data experiments demonstrate that, in joint auction scenarios, JTransNet significantly outperforms the considered baselines in terms of platform revenue.

Guogang Liao, Xuejian Li, Ze Wang et al.

Online advertising driven by auctions brings billions of dollars in revenue for social networking services and e-commerce platforms. GSP auctions, which are simple and easy to understand for advertisers, have almost become the benchmark for ad auction mechanisms in the industry. However, most GSP-based industrial practices assume that the user click only relies on the ad itself, which overlook the effect of external items, referred to as externalities. Recently, DNA has attempted to upgrade GSP with deep neural networks and models local externalities to some extent. However, it only considers set-level contexts from auctions and ignores the order and displayed position of ads, which is still suboptimal. Although VCG-based multi-slot auctions (e.g., VCG, WVCG) make it theoretically possible to model global externalities (e.g., the order and positions of ads and so on), they lack an efficient balance of both revenue and social welfare. In this paper, we propose novel auction mechanisms named Neural Multi-slot Auctions (NMA) to tackle the above-mentioned challenges. Specifically, we model the global externalities effectively with a context-aware list-wise prediction module to achieve better performance. We design a list-wise deep rank module to guarantee incentive compatibility in end-to-end learning. Furthermore, we propose an auxiliary loss for social welfare to effectively reduce the decline of social welfare while maximizing revenue. Experiment results on both offline large-scale datasets and online A/B tests demonstrate that NMA obtains higher revenue with balanced social welfare than other existing auction mechanisms (i.e., GSP, DNA, WVCG) in industrial practice, and we have successfully deployed NMA on Meituan food delivery platform.

Xiaowen Shi, Fan Yang, Ze Wang et al.

Re-ranking draws increased attention on both academics and industries, which rearranges the ranking list by modeling the mutual influence among items to better meet users' demands. Many existing re-ranking methods directly take the initial ranking list as input, and generate the optimal permutation through a well-designed context-wise model, which brings the evaluation-before-reranking problem. Meanwhile, evaluating all candidate permutations brings unacceptable computational costs in practice. Thus, to better balance efficiency and effectiveness, online systems usually use a two-stage architecture which uses some heuristic methods such as beam-search to generate a suitable amount of candidate permutations firstly, which are then fed into the evaluation model to get the optimal permutation. However, existing methods in both stages can be improved through the following aspects. As for generation stage, heuristic methods only use point-wise prediction scores and lack an effective judgment. As for evaluation stage, most existing context-wise evaluation models only consider the item context and lack more fine-grained feature context modeling. This paper presents a novel end-to-end re-ranking framework named PIER to tackle the above challenges which still follows the two-stage architecture and contains two mainly modules named FPSM and OCPM. We apply SimHash in FPSM to select top-K candidates from the full permutation based on user's permutation-level interest in an efficient way. Then we design a novel omnidirectional attention mechanism in OCPM to capture the context information in the permutation. Finally, we jointly train these two modules end-to-end by introducing a comparative learning loss. Offline experiment results demonstrate that PIER outperforms baseline models on both public and industrial datasets, and we have successfully deployed PIER on Meituan food delivery platform.

Wei Zhang, Pengye Zhang, Bo Zhang et al.

In the recommendation systems, there are multiple business domains to meet the diverse interests and needs of users, and the click-through rate(CTR) of each domain can be quite different, which leads to the demand for CTR prediction modeling for different business domains. The industry solution is to use domain-specific models or transfer learning techniques for each domain. The disadvantage of the former is that the data from other domains is not utilized by a single domain model, while the latter leverage all the data from different domains, but the fine-tuned model of transfer learning may trap the model in a local optimum of the source domain, making it difficult to fit the target domain. Meanwhile, significant differences in data quantity and feature schemas between different domains, known as domain shift, may lead to negative transfer in the process of transferring. To overcome these challenges, we propose the Collaborative Cross-Domain Transfer Learning Framework (CCTL). CCTL evaluates the information gain of the source domain on the target domain using a symmetric companion network and adjusts the information transfer weight of each source domain sample using the information flow network. This approach enables full utilization of other domain data while avoiding negative migration. Additionally, a representation enhancement network is used as an auxiliary task to preserve domain-specific features. Comprehensive experiments on both public and real-world industrial datasets, CCTL achieved SOTA score on offline metrics. At the same time, the CCTL algorithm has been deployed in Meituan, bringing 4.37% CTR and 5.43% GMV lift, which is significant to the business.

Guogang Liao, Xiaowen Shi, Ze Wang et al. · tsinghua

A mixed list of ads and organic items is usually displayed in feed and how to allocate the limited slots to maximize the overall revenue is a key problem. Meanwhile, modeling user preference with historical behavior is essential in recommendation and advertising (e.g., CTR prediction and ads allocation). Most previous works for user behavior modeling only model user's historical point-level positive feedback (i.e., click), which neglect the page-level information of feedback and other types of feedback. To this end, we propose Deep Page-level Interest Network (DPIN) to model the page-level user preference and exploit multiple types of feedback. Specifically, we introduce four different types of page-level feedback as input, and capture user preference for item arrangement under different receptive fields through the multi-channel interaction module. Through extensive offline and online experiments on Meituan food delivery platform, we demonstrate that DPIN can effectively model the page-level user preference and increase the revenue for the platform.

Wenjie Li, Shu-Tao Xia, Jiangke Fan et al.

Traditional vertical federated learning schema suffers from two main issues: 1) restricted applicable scope to overlapped samples and 2) high system challenge of real-time federated serving, which limits its application to advertising systems. To this end, we advocate a new practical learning setting, Semi-VFL (Vertical Semi-Federated Learning), for real-world industrial applications, where the learned model retains sufficient advantages of federated learning while supporting independent local serving. To achieve this goal, we propose the carefully designed Joint Privileged Learning framework (JPL) to i) alleviate the absence of the passive party's feature with federated equivalence imitation and ii) adapt to the heterogeneous full sample space with cross-branch rank alignment. Extensive experiments conducted on real-world advertising datasets validate the effectiveness of our method over baseline methods.

Huinan Sun, Guangliang Yu, Pengye Zhang et al.

Click-through rate (CTR) prediction aims to predict the probability that the user will click an item, which has been one of the key tasks in online recommender and advertising systems. In such systems, rich user behavior (viz. long- and short-term) has been proved to be of great value in capturing user interests. Both industry and academy have paid much attention to this topic and propose different approaches to modeling with long-term and short-term user behavior data. But there are still some unresolved issues. More specially, (1) rule and truncation based methods to extract information from long-term behavior are easy to cause information loss, and (2) single feedback behavior regardless of scenario to extract information from short-term behavior lead to information confusion and noise. To fill this gap, we propose a Graph based Long-term and Short-term interest Model, termed GLSM. It consists of a multi-interest graph structure for capturing long-term user behavior, a multi-scenario heterogeneous sequence model for modeling short-term information, then an adaptive fusion mechanism to fused information from long-term and short-term behaviors. Comprehensive experiments on real-world datasets, GLSM achieved SOTA score on offline metrics. At the same time, the GLSM algorithm has been deployed in our industrial application, bringing 4.9% CTR and 4.3% GMV lift, which is significant to the business.

Ze Wang, Guogang Liao, Xiaowen Shi et al. · tsinghua

Ads allocation, which involves allocating ads and organic items to limited slots in feed with the purpose of maximizing platform revenue, has become a research hotspot. Notice that, e-commerce platforms usually have multiple entrances for different categories and some entrances have few visits. Data from these entrances has low coverage, which makes it difficult for the agent to learn. To address this challenge, we propose Similarity-based Hybrid Transfer for Ads Allocation (SHTAA), which effectively transfers samples as well as knowledge from data-rich entrance to data-poor entrance. Specifically, we define an uncertainty-aware similarity for MDP to estimate the similarity of MDP for different entrances. Based on this similarity, we design a hybrid transfer method, including instance transfer and strategy transfer, to efficiently transfer samples and knowledge from one entrance to another. Both offline and online experiments on Meituan food delivery platform demonstrate that the proposed method could achieve better performance for data-poor entrance and increase the revenue for the platform.

Ze Wang, Guogang Liao, Xiaowen Shi et al. · tsinghua

With the recent prevalence of reinforcement learning (RL), there have been tremendous interests in utilizing RL for ads allocation in recommendation platforms (e.g., e-commerce and news feed sites). To achieve better allocation, the input of recent RL-based ads allocation methods is upgraded from point-wise single item to list-wise item arrangement. However, this also results in a high-dimensional space of state-action pairs, making it difficult to learn list-wise representations with good generalization ability. This further hinders the exploration of RL agents and causes poor sample efficiency. To address this problem, we propose a novel RL-based approach for ads allocation which learns better list-wise representations by leveraging task-specific signals on Meituan food delivery platform. Specifically, we propose three different auxiliary tasks based on reconstruction, prediction, and contrastive learning respectively according to prior domain knowledge on ads allocation. We conduct extensive experiments on Meituan food delivery platform to evaluate the effectiveness of the proposed auxiliary tasks. Both offline and online experimental results show that the proposed method can learn better list-wise representations and achieve higher revenue for the platform compared to the state-of-the-art baselines.

Wei Zhang, Ping Zhang, Jian Dong et al.

The effectiveness of ad creatives is greatly influenced by their visual appearance. Advertising platforms can generate ad creatives with different appearances by combining creative elements provided by advertisers. However, with the increasing number of ad creative elements, it becomes challenging to select a suitable combination from the countless possibilities. The industry's mainstream approach is to select individual creative elements independently, which often overlooks the importance of interaction between creative elements during the modeling process. In response, this paper proposes a Cross-Element Combinatorial Selection framework for multiple creative elements, termed CECS. In the encoder process, a cross-element interaction is adopted to dynamically adjust the expression of a single creative element based on the current candidate creatives. In the decoder process, the creative combination problem is transformed into a cascade selection problem of multiple creative elements. A pointer mechanism with a cascade design is used to model the associations among candidates. Comprehensive experiments on real-world datasets show that CECS achieved the SOTA score on offline metrics. Moreover, the CECS algorithm has been deployed in our industrial application, resulting in a significant 6.02% CTR and 10.37% GMV lift, which is beneficial to the business.

Yongcan Yu, Lingxiao He, Jian Liang et al.

Test-time reinforcement learning (TTRL) always adapts models at inference time via pseudo-labeling, leaving it vulnerable to spurious optimization signals from label noise. Through an empirical study, we observe that responses with medium consistency form an ambiguity region and constitute the primary source of reward noise. Crucially, we find that such spurious signals can be even amplified through group-relative advantage estimation. Motivated by these findings, we propose a unified framework, Debiased and Denoised test-time Reinforcement Learning (DDRL), to mitigate spurious signals. Concretely, DDRL first applies a frequency-based sampling strategy to exclude ambiguous samples while maintaining a balanced set of positive and negative examples. It then adopts a debiased advantage estimation with fixed advantages, removing the bias introduced by group-relative policy optimization. Finally, DDRL incorporates a consensus-based off-policy refinement stage, which leverages the rejection-sampled dataset to enable efficient and stable model updates. Experiments on three large language models across multiple mathematical reasoning benchmarks demonstrate that DDRL consistently outperforms existing TTRL baselines. The code will soon be released at https://github.com/yuyongcan/DDRL.

Jian Dong, Yisong Yu, Yapeng Zhang et al.

User behaviors on an e-commerce app not only contain different kinds of feedback on items but also sometimes imply the cognitive clue of the user's decision-making. For understanding the psychological procedure behind user decisions, we present the behavior path and propose to match the user's current behavior path with historical behavior paths to predict user behaviors on the app. Further, we design a deep neural network for behavior path matching and solve three difficulties in modeling behavior paths: sparsity, noise interference, and accurate matching of behavior paths. In particular, we leverage contrastive learning to augment user behavior paths, provide behavior path self-activation to alleviate the effect of noise, and adopt a two-level matching mechanism to identify the most appropriate candidate. Our model shows excellent performance on two real-world datasets, outperforming the state-of-the-art CTR model. Moreover, our model has been deployed on the Meituan food delivery platform and has accumulated 1.6% improvement in CTR and 1.8% improvement in advertising revenue.

Qian Lin, Bo Tang, Zifan Wu et al.

Aiming at promoting the safe real-world deployment of Reinforcement Learning (RL), research on safe RL has made significant progress in recent years. However, most existing works in the literature still focus on the online setting where risky violations of the safety budget are likely to be incurred during training. Besides, in many real-world applications, the learned policy is required to respond to dynamically determined safety budgets (i.e., constraint threshold) in real time. In this paper, we target at the above real-time budget constraint problem under the offline setting, and propose Trajectory-based REal-time Budget Inference (TREBI) as a novel solution that models this problem from the perspective of trajectory distribution and solves it through diffusion model planning. Theoretically, we prove an error bound of the estimation on the episodic reward and cost under the offline setting and thus provide a performance guarantee for TREBI. Empirical results on a wide range of simulation tasks and a real-world large-scale advertising application demonstrate the capability of TREBI in solving real-time budget constraint problems under offline settings.

Qingyu Zhang, Chunlei Xin, Xuanang Chen et al.

Goal-driven persuasive dialogue, exemplified by applications like telemarketing, requires sophisticated multi-turn planning and strict factual faithfulness, which remains a significant challenge for even state-of-the-art Large Language Models (LLMs). A lack of task-specific data often limits previous works, and direct LLM application suffers from strategic brittleness and factual hallucination. In this paper, we first construct and release TeleSalesCorpus, the first real-world-grounded dialogue dataset for this domain. We then propose AI-Salesman, a novel framework featuring a dual-stage architecture. For the training stage, we design a Bayesian-supervised reinforcement learning algorithm that learns robust sales strategies from noisy dialogues. For the inference stage, we introduce the Dynamic Outline-Guided Agent (DOGA), which leverages a pre-built script library to provide dynamic, turn-by-turn strategic guidance. Moreover, we design a comprehensive evaluation framework that combines fine-grained metrics for key sales skills with the LLM-as-a-Judge paradigm. Experimental results demonstrate that our proposed AI-Salesman significantly outperforms baseline models in both automatic metrics and comprehensive human evaluations, showcasing its effectiveness in complex persuasive scenarios.

Xiang Li, Shuwei Chen, Jian Dong et al.

Click-through rate (CTR) prediction is crucial in recommendation and online advertising systems. Existing methods usually model user behaviors, while ignoring the informative context which influences the user to make a click decision, e.g., click pages and pre-ranking candidates that inform inferences about user interests, leading to suboptimal performance. In this paper, we propose a Decision-Making Context Interaction Network (DCIN), which deploys a carefully designed Context Interaction Unit (CIU) to learn decision-making contexts and thus benefits CTR prediction. In addition, the relationship between different decision-making context sources is explored by the proposed Adaptive Interest Aggregation Unit (AIAU) to improve CTR prediction further. In the experiments on public and industrial datasets, DCIN significantly outperforms the state-of-the-art methods. Notably, the model has obtained the improvement of CTR+2.9%/CPM+2.1%/GMV+1.5% for online A/B testing and served the main traffic of Meituan Waimai advertising system.

Weide Liu, Xingxing Wang, Lu Wang et al.

In this paper, we introduce a novel Gaussian mixture based evidential learning solution for robust stereo matching. Diverging from previous evidential deep learning approaches that rely on a single Gaussian distribution, our framework posits that individual image data adheres to a mixture-of-Gaussian distribution in stereo matching. This assumption yields more precise pixel-level predictions and more accurately mirrors the real-world image distribution. By further employing the inverse-Gamma distribution as an intermediary prior for each mixture component, our probabilistic model achieves improved depth estimation compared to its counterpart with the single Gaussian and effectively captures the model uncertainty, which enables a strong cross-domain generation ability. We evaluated our method for stereo matching by training the model using the Scene Flow dataset and testing it on KITTI 2015 and Middlebury 2014. The experiment results consistently show that our method brings improvements over the baseline methods in a trustworthy manner. Notably, our approach achieved new state-of-the-art results on both the in-domain validated data and the cross-domain datasets, demonstrating its effectiveness and robustness in stereo matching tasks.

Zhaoqi Zhang, Jiaming Deng, Miao Xie et al.

Guaranteed display advertising is crucial for platform monetization, yet existing methods often operate under a single-slot assumption, limiting their ability to optimize allocation across multi-slot page views. In this paper, we propose a novel joint optimization framework for multi-slot GD allocation, addressing key challenges such as slot-level redundancy, contract imbalance, and exposure concentration. Our approach formulates the allocation as an offline bipartite matching problem with a contract roulette mechanism for slot exclusivity and Page View constraints for impression control, and incorporates a scalable allocation optimization algorithm for efficient large-scale deployment. Extensive online tests on the Meituan advertising platform demonstrate that our method significantly improves merchant ROI, platform revenue efficiency, and contract fulfillment robustness. Specifically, online A/B tests show a 28.99% increase in Average Revenue Per User under 70% traffic, and DID analysis further indicates improved contract stability, demonstrating the strong applicability and effectiveness of our framework in real-world advertising deployments.

Xingyu Hu, Kai Zhang, Jiancan Wu et al.

In large language model (LLM)-based recommendation systems, direct preference optimization (DPO) effectively aligns recommendations with user preferences, requiring multi-negative objective functions to leverage abundant implicit-feedback negatives and sharpen preference boundaries. However, our empirical analyses reveal a counterintuitive phenomenon, preference optimization collapse, where increasing the number of negative samples can lead to performance degradation despite a continuously decreasing training loss. We further theoretically demonstrate that this collapse arises from gradient suppression, caused by the dominance of easily discriminable negatives over boundary-critical negatives that truly define user preference boundaries. As a result, boundary-relevant signals are under-optimized, weakening the model's decision boundary. Motivated by these observations, we propose DynamicPO (Dynamic Preference Optimization), a lightweight and plug-and-play framework comprising two adaptive mechanisms: Dynamic Boundary Negative Selection, which identifies and prioritizes informative negatives near the model's decision boundary, and Dual-Margin Dynamic beta Adjustment, which calibrates optimization strength per sample according to boundary ambiguity. Extensive experiments on three public datasets show that DynamicPO effectively prevents optimization collapse and improves recommendation accuracy on multi-negative preference optimization methods, with negligible computational overhead. Our code and datasets are available at https://github.com/xingyuHuxingyu/DynamicPO.

Shuwei Chen, Xiang Li, Jian Dong et al.

Click-through rate (CTR) prediction plays a pivotal role in the success of recommendations. Inspired by the recent thriving of language models (LMs), a surge of works improve prediction by organizing user behavior data in a \textbf{textual} format and using LMs to understand user interest at a semantic level. While promising, these works have to truncate the textual data to reduce the quadratic computational overhead of self-attention in LMs. However, it has been studied that long user behavior data can significantly benefit CTR prediction. In addition, these works typically condense user diverse interests into a single feature vector, which hinders the expressive capability of the model. In this paper, we propose a \textbf{T}extual \textbf{B}ehavior-based \textbf{I}nterest Chunking \textbf{N}etwork (TBIN), which tackles the above limitations by combining an efficient locality-sensitive hashing algorithm and a shifted chunk-based self-attention. The resulting user diverse interests are dynamically activated, producing user interest representation towards the target item. Finally, the results of both offline and online experiments on real-world food recommendation platform demonstrate the effectiveness of TBIN.

Xiaowen Shi, Ze Wang, Yuanying Cai et al.

Nowadays, the mainstream approach in position allocation system is to utilize a reinforcement learning model to allocate appropriate locations for items in various channels and then mix them into the feed. There are two types of data employed to train reinforcement learning (RL) model for position allocation, named strategy data and random data. Strategy data is collected from the current online model, it suffers from an imbalanced distribution of state-action pairs, resulting in severe overestimation problems during training. On the other hand, random data offers a more uniform distribution of state-action pairs, but is challenging to obtain in industrial scenarios as it could negatively impact platform revenue and user experience due to random exploration. As the two types of data have different distributions, designing an effective strategy to leverage both types of data to enhance the efficacy of the RL model training has become a highly challenging problem. In this study, we propose a framework named Multi-Distribution Data Learning (MDDL) to address the challenge of effectively utilizing both strategy and random data for training RL models on mixed multi-distribution data. Specifically, MDDL incorporates a novel imitation learning signal to mitigate overestimation problems in strategy data and maximizes the RL signal for random data to facilitate effective learning. In our experiments, we evaluated the proposed MDDL framework in a real-world position allocation system and demonstrated its superior performance compared to the previous baseline. MDDL has been fully deployed on the Meituan food delivery platform and currently serves over 300 million users.

Yimin Lv, Shuli Wang, Beihong Jin et al.

User behavior sequence modeling plays a significant role in Click-Through Rate (CTR) prediction on e-commerce platforms. Except for the interacted items, user behaviors contain rich interaction information, such as the behavior type, time, location, etc. However, so far, the information related to user behaviors has not yet been fully exploited. In the paper, we propose the concept of a situation and situational features for distinguishing interaction behaviors and then design a CTR model named Deep Situation-Aware Interaction Network (DSAIN). DSAIN first adopts the reparameterization trick to reduce noise in the original user behavior sequences. Then it learns the embeddings of situational features by feature embedding parameterization and tri-directional correlation fusion. Finally, it obtains the embedding of behavior sequence via heterogeneous situation aggregation. We conduct extensive offline experiments on three real-world datasets. Experimental results demonstrate the superiority of the proposed DSAIN model. More importantly, DSAIN has increased the CTR by 2.70\%, the CPM by 2.62\%, and the GMV by 2.16\% in the online A/B test. Now, DSAIN has been deployed on the Meituan food delivery platform and serves the main traffic of the Meituan takeout app.

Jiangli Shao, Kaifu Zheng, Hao Fang et al.

Modeling long-term user interests with massive historical user behaviors enhances click-through rate (CTR) prediction performance in advertising and recommendation systems. Typically, a two-stage framework is widely adopted, where a general search unit (GSU) first retrieves top-$k$ relevant behaviors towards the target item, and an exact search unit (ESU) generates interest features via tailored attention. However, current target-centered GSU would ignore other latent user interests, leading to incomplete and biased interest features. Additionally, the matching-based retrieval process in GSUs depends on the pairwise similarity score between target item and each historical behavior, which not only becomes time-consuming for online services as user behaviors continue to grow, but also overlooks the interaction information among user behaviors. To combat these problems, we propose a \textbf{Gen}erative \textbf{L}ong-term user \textbf{I}nterest model named GenLI for CTR prediction. GenLI consists of an interest generation module (IGM), a behavior retrieval module (BRM), and an interest fusion module (IFM). The IGM generates multiple interest distributions to indicate different aspects of real-time user interests, which is target-independent and incorporates interaction information among behaviors, ensuring complete and diverse interest features. The BRM selects related behaviors via a simple lookup operation, reducing the time complexity for weighting each behavior to $O(1)$. Finally, the IFM uses delicate gating mechanisms to generate interest features. Based on the generation process, GenLI improves the diversity of user interests and avoids complex matching-based behavioral retrieval, achieving a better balance between accuracy and efficiency for CTR prediction.

Shuli Wang, Junwei Yin, Changhao Li et al.

Semantic IDs (SIDs) define the generation space of generative recommendation and directly determine its personalization ceiling. However, existing tokenizers are trained independently with retrieval objectives, leaving personalization signals fully decoupled from the SID construction process -- a fundamental gap that causes generative retrieval to persistently lag behind discriminative ranking. In this paper, we rethink the essence of SIDs: \emph{ranking seeks argmax in item space while retrieval seeks argmax in token space; both are the same problem solved at different granularities.} Based on this insight, we propose \DIG (\textbf{D}iscrimination \textbf{I}s \textbf{G}eneration), which embeds the tokenizer inside a discriminative ranking model for end-to-end training -- the ranker naturally becomes a retrieval model, yielding two models from a single training run. \DIG is organized around a \emph{feature assignment taxonomy}: item-intrinsic static features are encoded into SIDs, user-item cross features (u2i) implicitly drive codebook boundaries toward recommendation decision boundaries during training, and an MLP$_\mathrm{u2t}$ distillation module approximates u2i at the token level for inference. Experiments on three public benchmarks and two industrial datasets demonstrate that \DIG simultaneously improves ranking, retrieval, and unified retrieval-ranking quality.

Shuo Lu, Jianjie Cheng, Yinuo Xu et al.

Multimodal large language models (MLLMs) have achieved strong performance on perception-oriented tasks, yet their ability to perform mathematical spatial reasoning, defined as the capacity to parse and manipulate two- and three-dimensional relations, remains unclear. Humans easily solve textbook-style spatial reasoning problems with over 95\% accuracy, but we find that most leading MLLMs fail to reach even 60\% on the same tasks. This striking gap highlights spatial reasoning as a fundamental weakness of current models. To investigate this gap, we present MathSpatial, a unified framework for evaluating and improving spatial reasoning in MLLMs. MathSpatial includes three complementary components: (i) MathSpatial-Bench, a benchmark of 2K problems across three categories and eleven subtypes, designed to isolate reasoning difficulty from perceptual noise; (ii) MathSpatial-Corpus, a training dataset of 8K additional problems with verified solutions; and (iii) MathSpatial-SRT, which models reasoning as structured traces composed of three atomic operations--Correlate, Constrain, and Infer. Experiments show that fine-tuning Qwen2.5-VL-7B on MathSpatial achieves competitive accuracy while reducing tokens by 25\%. MathSpatial provides the first large-scale resource that disentangles perception from reasoning, enabling precise measurement and comprehensive understanding of mathematical spatial reasoning in MLLMs.

Boyu Chen, Tai Guo, Weiyu Cui et al.

Multimodal retrieval models are becoming increasingly important in scenarios such as food delivery, where rich multimodal features can meet diverse user needs and enable precise retrieval. Mainstream approaches typically employ a dual-tower architecture between queries and items, and perform joint optimization of intra-tower and inter-tower tasks. However, we observe that joint optimization often leads to certain modalities dominating the training process, while other modalities are neglected. In addition, inconsistent training speeds across modalities can easily result in the one-epoch problem. To address these challenges, we propose a staged pretraining strategy, which guides the model to focus on specialized tasks at each stage, enabling it to effectively attend to and utilize multimodal features, and allowing flexible control over the training process at each stage to avoid the one-epoch problem. Furthermore, to better utilize the semantic IDs that compress high-dimensional multimodal embeddings, we design both generative and discriminative tasks to help the model understand the associations between SIDs, queries, and item features, thereby improving overall performance. Extensive experiments on large-scale real-world Meituan data demonstrate that our method achieves improvements of 3.80%, 2.64%, and 2.17% on R@5, R@10, and R@20, and 5.10%, 4.22%, and 2.09% on N@5, N@10, and N@20 compared to mainstream baselines. Online A/B testing on the Meituan platform shows that our approach achieves a 1.12% increase in revenue and a 1.02% increase in click-through rate, validating the effectiveness and superiority of our method in practical applications.

Shuli Wang, Junwei Yin, Changhao Li et al.

Scaling industrial recommender models has followed two parallel paradigms: \textbf{sample information scaling} -- enriching the information content of each training sample through deeper and longer behavior sequences -- and \textbf{model capacity scaling} -- unifying sequence modeling and feature interaction within a single Transformer backbone. However, these two paradigms still face two structural limitations. Firstly, sample information scaling methods encode only a subset of each historical interaction into the sequence token, leaving the majority of the original sample context unexploited and precluding the modeling of sample-level, time-varying features. Secondly, model capacity scaling methods are inherently constrained by the structural heterogeneity between sequential and non-sequential features, preventing the model from fully realizing its representational capacity. To address these issues, we propose \textbf{SIF} (\emph{Sample Is Feature}), which encodes each historical Raw Sample directly into the sequence token -- maximally preserving sample information while simultaneously resolving the heterogeneity between sequential and non-sequential features. SIF consists of two key components. The \textbf{Sample Tokenizer} quantizes each historical Raw Sample into a Token Sample via hierarchical group-adaptive quantization (HGAQ), enabling full sample-level context to be incorporated into the sequence efficiently. The \textbf{SIF-Mixer} then performs deep feature interaction over the homogeneous sample representations via token-level and sample-level mixing, fully unleashing the model's representational capacity. Extensive experiments on a large-scale industrial dataset validate SIF's effectiveness, and we have successfully deployed SIF on the Meituan food delivery platform.

Zifan Wu, Bo Tang, Qian Lin et al.

Primal-dual safe RL methods commonly perform iterations between the primal update of the policy and the dual update of the Lagrange Multiplier. Such a training paradigm is highly susceptible to the error in cumulative cost estimation since this estimation serves as the key bond connecting the primal and dual update processes. We show that this problem causes significant underestimation of cost when using off-policy methods, leading to the failure to satisfy the safety constraint. To address this issue, we propose conservative policy optimization, which learns a policy in a constraint-satisfying area by considering the uncertainty in cost estimation. This improves constraint satisfaction but also potentially hinders reward maximization. We then introduce local policy convexification to help eliminate such suboptimality by gradually reducing the estimation uncertainty. We provide theoretical interpretations of the joint coupling effect of these two ingredients and further verify them by extensive experiments. Results on benchmark tasks show that our method not only achieves an asymptotic performance comparable to state-of-the-art on-policy methods while using much fewer samples, but also significantly reduces constraint violation during training. Our code is available at https://github.com/ZifanWu/CAL.

Yinuo Xu, Shuo Lu, Jianjie Cheng et al.

Deep Research agents tackle knowledge-intensive tasks through multi-round retrieval and decision-oriented generation. While reinforcement learning (RL) has been shown to improve performance in this paradigm, its contributions remain underexplored. To fully understand the role of RL, we conduct a systematic study along three decoupled dimensions: prompt template, reward function, and policy optimization. Our study reveals that: 1) the Fast Thinking template yields greater stability and better performance than the Slow Thinking template used in prior work; 2) the F1-based reward underperforms the EM due to training collapse driven by answer avoidance; this can be mitigated by incorporating action-level penalties, ultimately surpassing EM; 3) REINFORCE outperforms PPO while requiring fewer search actions, whereas GRPO shows the poorest stability among policy optimization methods. Building on these insights, we then introduce Search-R1++, a strong baseline that improves the performance of Search-R1 from 0.403 to 0.442 (Qwen2.5-7B) and 0.289 to 0.331 (Qwen2.5-3B). We hope that our findings can pave the way for more principled and reliable RL training strategies in Deep Research systems.

Guoxiao Zhang, Tan Qu, Ao Li et al.

Reranking improves recommendation quality by modeling item interactions. However, existing methods often decouple ranking and reranking, leading to weak listwise evaluation models that suffer from combinatorial sparsity and limited representational power under strict latency constraints. In this paper, we propose RIA (Ranking-Infused Architecture), a unified, end-to-end framework that seamlessly integrates pointwise and listwise evaluation. RIA introduces four key components: (1) the User and Candidate DualTransformer (UCDT) for fine-grained user-item-context modeling; (2) the Context-aware User History and Target (CUHT) module for position-sensitive preference learning; (3) the Listwise Multi-HSTU (LMH) module to capture hierarchical item dependencies; and (4) the Embedding Cache (EC) module to bridge efficiency and effectiveness during inference. By sharing representations across ranking and reranking, RIA enables rich contextual knowledge transfer while maintaining low latency. Extensive experiments show that RIA outperforms state-of-the-art models on both public and industrial datasets, achieving significant gains in AUC and LogLoss. Deployed in Meituan advertising system, RIA yields a +1.69% improvement in Click-Through Rate (CTR) and a +4.54% increase in Cost Per Mille (CPM) in online A/B tests.

Guoxiao Zhang, Ao Li, Tan Qu et al.

Online platforms usually suffer from user experience degradation due to near-duplicate items with similar visuals and text. While Multimodal Large Language Models (MLLMs) enable multimodal embedding, existing methods treat representations as black boxes, ignoring structural relationships (e.g., primary vs. auxiliary elements), leading to local structural collapse problem. To address this, inspired by Feature Integration Theory (FIT), we propose FITRep, the first attention-guided, white-box item representation framework for fine-grained item deduplication. FITRep consists of: (1) Concept Hierarchical Information Extraction (CHIE), using MLLMs to extract hierarchical semantic concepts; (2) Structure-Preserving Dimensionality Reduction (SPDR), an adaptive UMAP-based method for efficient information compression; and (3) FAISS-Based Clustering (FBC), a FAISS-based clustering that assigns each item a unique cluster id using FAISS. Deployed on Meituan's advertising system, FITRep achieves +3.60% CTR and +4.25% CPM gains in online A/B tests, demonstrating both effectiveness and real-world impact.

Shuli Wang, Changhao Li, Ke Fan et al.

In modern multi-stage recommendation systems, reranking plays a critical role by modeling contextual information. Due to inherent challenges such as the combinatorial space complexity, an increasing number of methods adopt the generative paradigm: the generator produces the optimal list during inference, while an evaluator guides the generator's optimization during the training phase. However, these methods still face two problems. Firstly, these generators fail to produce optimal generation results due to the lack of both local and global perspectives, regardless of whether the generation strategy is autoregressive or non-autoregressive. Secondly, the goal inconsistency problem between the generator and the evaluator during training complicates the guidance signal and leading to suboptimal performance. To address these issues, we propose the \textbf{N}ext-\textbf{S}cale \textbf{G}eneration \textbf{R}eranking (NSGR), a tree-based generative framework. Specifically, we introduce a next-scale generator (NSG) that progressively expands a recommendation list from user interests in a coarse-to-fine manner, balancing global and local perspectives. Furthermore, we design a multi-scale neighbor loss, which leverages a tree-based multi-scale evaluator (MSE) to provide scale-specific guidance to the NSG at each scale. Extensive experiments on public and industrial datasets validate the effectiveness of NSGR. And NSGR has been successfully deployed on the Meituan food delivery platform.

Changhao Li, Junwei Yin, Zhilin Zeng et al.

Generative recommendation (GR) has recently emerged as a promising paradigm for industrial recommendations. GR leverages Semantic IDs (SIDs) to reduce the encoding-decoding space and employs the Next Token Prediction (NTP) framework to explore scaling laws. However, existing GR methods suffer from two critical issues: (1) a \textbf{seesaw phenomenon} in multi-business scenarios arises due to NTP's inability to capture complex cross-business behavioral patterns; and (2) a unified SID space causes \textbf{representation confusion} by failing to distinguish distinct semantic information across businesses. To address these issues, we propose Multi-Business Generative Recommendation (MBGR), the first GR framework tailored for multi-business scenarios. Our framework comprises three key components. First, we design a Business-aware semantic ID (BID) module that preserves semantic integrity via domain-aware tokenization. Then, we introduce a Multi-Business Prediction (MBP) structure to provide business-specific prediction capabilities. Furthermore, we develop a Label Dynamic Routing (LDR) module that transforms sparse multi-business labels into dense labels to further enhance the multi-business generation capability. Extensive offline and online experiments on Meituan's food delivery platform validate MBGR's effectiveness, and we have successfully deployed it in production.

Xuejian Li, Ze Wang, Bingqi Zhu et al.

E-commerce platforms usually present an ordered list, mixed with several organic items and an advertisement, in response to each user's page view request. This list, the outcome of ad auction and allocation processes, directly impacts the platform's ad revenue and gross merchandise volume (GMV). Specifically, the ad auction determines which ad is displayed and the corresponding payment, while the ad allocation decides the display positions of the advertisement and organic items. The prevalent methods of segregating the ad auction and allocation into two distinct stages face two problems: 1) Ad auction does not consider externalities, such as the influence of actual display position and context on ad Click-Through Rate (CTR); 2) The ad allocation, which utilizes the auction-winning ad's payment to determine the display position dynamically, fails to maintain incentive compatibility (IC) for the advertisement. For instance, in the auction stage employing the traditional Generalized Second Price (GSP) , even if the winning ad increases its bid, its payment remains unchanged. This implies that the advertisement cannot secure a better position and thus loses the opportunity to achieve higher utility in the subsequent ad allocation stage. Previous research often focused on one of the two stages, neglecting the two-stage problem, which may result in suboptimal outcomes...

Jiahong Zhou, Shunhui Mao, Guoliang Yang et al.

Recommender systems aim to recommend the most suitable items to users from a large number of candidates. Their computation cost grows as the number of user requests and the complexity of services (or models) increases. Under the limitation of computation resources (CRs), how to make a trade-off between computation cost and business revenue becomes an essential question. The existing studies focus on dynamically allocating CRs in queue truncation scenarios (i.e., allocating the size of candidates), and formulate the CR allocation problem as an optimization problem with constraints. Some of them focus on single-phase CR allocation, and others focus on multi-phase CR allocation but introduce some assumptions about queue truncation scenarios. However, these assumptions do not hold in other scenarios, such as retrieval channel selection and prediction model selection. Moreover, existing studies ignore the state transition process of requests between different phases, limiting the effectiveness of their approaches. This paper proposes a Reinforcement Learning (RL) based Multi-Phase Computation Allocation approach (RL-MPCA), which aims to maximize the total business revenue under the limitation of CRs. RL-MPCA formulates the CR allocation problem as a Weakly Coupled MDP problem and solves it with an RL-based approach. Specifically, RL-MPCA designs a novel deep Q-network to adapt to various CR allocation scenarios, and calibrates the Q-value by introducing multiple adaptive Lagrange multipliers (adaptive-$λ$) to avoid violating the global CR constraints. Finally, experiments on the offline simulation environment and online real-world recommender system validate the effectiveness of our approach.

Shuli Wang, Xue Wei, Senjie Kou et al.

Reranking plays a crucial role in modern multi-stage recommender systems by rearranging the initial ranking list. Due to the inherent challenges of combinatorial search spaces, some current research adopts an evaluator-generator paradigm, with a generator generating feasible sequences and an evaluator selecting the best sequence based on the estimated list utility. However, these methods still face two issues. Firstly, due to the goal inconsistency problem between the evaluator and generator, the generator tends to fit the local optimal solution of exposure distribution rather than combinatorial space optimization. Secondly, the strategy of generating target items one by one is difficult to achieve optimality because it ignores the information of subsequent items. To address these issues, we propose a utilizing Neighbor Lists model for Generative Reranking (NLGR), which aims to improve the performance of the generator in the combinatorial space. NLGR follows the evaluator-generator paradigm and improves the generator's training and generating methods. Specifically, we use neighbor lists in combination space to enhance the training process, making the generator perceive the relative scores and find the optimization direction. Furthermore, we propose a novel sampling-based non-autoregressive generation method, which allows the generator to jump flexibly from the current list to any neighbor list. Extensive experiments on public and industrial datasets validate NLGR's effectiveness and we have successfully deployed NLGR on the Meituan food delivery platform.

Hao Wang, Bo Tang, Chi Harold Liu et al.

Online display advertising platforms service numerous advertisers by providing real-time bidding (RTB) for the scale of billions of ad requests every day. The bidding strategy handles ad requests cross multiple channels to maximize the number of clicks under the set financial constraints, i.e., total budget and cost-per-click (CPC), etc. Different from existing works mainly focusing on single channel bidding, we explicitly consider cross-channel constrained bidding with budget allocation. Specifically, we propose a hierarchical offline deep reinforcement learning (DRL) framework called ``HiBid'', consisted of a high-level planner equipped with auxiliary loss for non-competitive budget allocation, and a data augmentation enhanced low-level executor for adaptive bidding strategy in response to allocated budgets. Additionally, a CPC-guided action selection mechanism is introduced to satisfy the cross-channel CPC constraint. Through extensive experiments on both the large-scale log data and online A/B testing, we confirm that HiBid outperforms six baselines in terms of the number of clicks, CPC satisfactory ratio, and return-on-investment (ROI). We also deploy HiBid on Meituan advertising platform to already service tens of thousands of advertisers every day.

Yongcan Yu, Lingxiao He, Shuo Lu et al.

Recent advances in vision-language models (VLMs) reasoning have been largely attributed to the rise of reinforcement Learning (RL), which has shifted the community's focus away from the supervised fine-tuning (SFT) paradigm. Many studies suggest that introducing the SFT stage not only fails to improve reasoning ability but may also negatively impact model training. In this study, we revisit this RL-centric belief through a systematic and controlled comparison of SFT and RL on VLM Reasoning. Using identical data sources, we find that the relative effectiveness of SFT and RL is conditional and strongly influenced by model capacity, data scale, and data distribution. Contrary to common assumptions, our findings show that SFT plays a crucial role across several scenarios: (1) Effectiveness for weaker models. SFT more reliably elicits reasoning capabilities in smaller or weaker VLMs. (2) Data efficiency. SFT with only 2K achieves comparable or better reasoning performance to RL with 20K. (3) Cross-modal transferability. SFT demonstrates stronger generalization across modalities. Moreover, we identify a pervasive issue of deceptive rewards, where higher rewards fail to correlate with better reasoning accuracy in RL. These results challenge the prevailing "RL over SFT" narrative. They highlight that the role of SFT may have been underestimated and support a more balanced post-training pipeline in which SFT and RL function as complementary components.

Shuwei Chen, Jiajun Cui, Zhengqi Xu et al.

Click-through rate (CTR) prediction, which models behavior sequence and non-sequential features (e.g., user/item profiles or cross features) to infer user interest, underpins industrial recommender systems. However, most methods face three forms of heterogeneity that degrade predictive performance: (i) Feature Heterogeneity persists when limited sequence side features provide less granular interest representation compared to extensive non-sequential features, thereby impairing sequence modeling performance; (ii) Context Heterogeneity arises because a user's interest in an item will be influenced by other items, yet point-wise prediction neglects cross-item interaction context from the entire item set; (iii) Architecture Heterogeneity stems from the fragmented integration of specialized network modules, which compounds the model's effectiveness, efficiency and scalability in industrial deployments. To tackle the above limitations, we propose HoMer, a Homogeneous-Oriented TransforMer for modeling sequential and set-wise contexts. First, we align sequence side features with non-sequential features for accurate sequence modeling and fine-grained interest representation. Second, we shift the prediction paradigm from point-wise to set-wise, facilitating cross-item interaction in a highly parallel manner. Third, HoMer's unified encoder-decoder architecture achieves dual optimization through structural simplification and shared computation, ensuring computational efficiency while maintaining scalability with model size. Without arduous modification to the prediction pipeline, HoMer successfully scales up and outperforms our industrial baseline by 0.0099 in the AUC metric, and enhances online business metrics like CTR/RPM by 1.99%/2.46%. Additionally, HoMer saves 27% of GPU resources via preliminary engineering optimization, further validating its superiority and practicality.

Chenlu Ding, Daoxuan Liu, Jiancan Wu et al.

Recommendation systems leverage user interaction data to suggest relevant items while filtering out irrelevant (negative) ones. The rise of large language models (LLMs) has garnered increasing attention for their potential in recommendation tasks. However, existing methods for optimizing LLM-based recommenders face challenges in effectively utilizing negative samples. Simply integrating large numbers of negative samples can improve ranking accuracy and mitigate popularity bias but often leads to increased computational overhead and memory costs. Additionally, current approaches fail to account for the varying informativeness of negative samples, leading to suboptimal optimization performance. To address these issues, we propose NAPO (\textbf{N}egative-\textbf{A}ware \textbf{P}reference \textbf{O}ptimization), an enhanced framework for preference optimization in LLM-based recommendation. NAPO introduces two key innovations: (1) in-batch negative sharing, which expands the pool of negative samples without additional memory overhead, and (2) dynamic reward margin adjustment, which adapts model updates based on the confidence of negative samples. Extensive experiments on three public datasets demonstrate that NAPO outperforms existing methods in both recommendation accuracy and popularity bias reduction.

Yinqiu Huang, Hao Ma, Wenshuai Chen et al.

Bid shading plays a crucial role in Real-Time Bidding~(RTB) by adaptively adjusting the bid to avoid advertisers overspending. Existing mainstream two-stage methods, which first model bid landscapes and then optimize surplus using operations research techniques, are constrained by unimodal assumptions that fail to adapt for non-convex surplus curves and are vulnerable to cascading errors in sequential workflows. Additionally, existing discretization models of continuous values ignore the dependence between discrete intervals, reducing the model's error correction ability, while sample selection bias in bidding scenarios presents further challenges for prediction. To address these issues, this paper introduces Generative Bid Shading~(GBS), which comprises two primary components: (1) an end-to-end generative model that utilizes an autoregressive approach to generate shading ratios by stepwise residuals, capturing complex value dependencies without relying on predefined priors; and (2) a reward preference alignment system, which incorporates a channel-aware hierarchical dynamic network~(CHNet) as the reward model to extract fine-grained features, along with modules for surplus optimization and exploration utility reward alignment, ultimately optimizing both short-term and long-term surplus using group relative policy optimization~(GRPO). Extensive experiments on both offline and online A/B tests validate GBS's effectiveness. Moreover, GBS has been deployed on the Meituan DSP platform, serving billions of bid requests daily.

Yu Lei, Jiayang Zhao, Yilei Zhao et al.

Modern auto-bidding systems are required to balance overall performance with diverse advertiser goals and real-world constraints, reflecting the dynamic and evolving needs of the industry. Recent advances in conditional generative models, such as transformers and diffusers, have enabled direct trajectory generation tailored to advertiser preferences, offering a promising alternative to traditional Markov Decision Process-based methods. However, these generative methods face significant challenges, such as the distribution shift between offline and online environments, limited exploration of the action space, and the necessity to meet constraints like marginal Cost-per-Mille (CPM) and Return on Investment (ROI). To tackle these challenges, we propose GRAD (Generative Reward-driven Ad-bidding with Mixture-of-Experts), a scalable foundation model for auto-bidding that combines an Action-Mixture-of-Experts module for diverse bidding action exploration with the Value Estimator of Causal Transformer for constraint-aware optimization. Extensive offline and online experiments demonstrate that GRAD significantly enhances platform revenue, highlighting its effectiveness in addressing the evolving and diverse requirements of modern advertisers. Furthermore, GRAD has been implemented in multiple marketing scenarios at Meituan, one of the world's largest online food delivery platforms, leading to a 2.18% increase in Gross Merchandise Value (GMV) and 10.68% increase in ROI.

Guogang Liao, Ze Wang, Xiaoxu Wu et al.

E-commerce platforms usually display a mixed list of ads and organic items in feed. One key problem is to allocate the limited slots in the feed to maximize the overall revenue as well as improve user experience, which requires a good model for user preference. Instead of modeling the influence of individual items on user behaviors, the arrangement signal models the influence of the arrangement of items and may lead to a better allocation strategy. However, most of previous strategies fail to model such a signal and therefore result in suboptimal performance. In addition, the percentage of ads exposed (PAE) is an important indicator in ads allocation. Excessive PAE hurts user experience while too low PAE reduces platform revenue. Therefore, how to constrain the PAE within a certain range while keeping personalized recommendation under the PAE constraint is a challenge. In this paper, we propose Cross Deep Q Network (Cross DQN) to extract the crucial arrangement signal by crossing the embeddings of different items and modeling the crossed sequence by multi-channel attention. Besides, we propose an auxiliary loss for batch-level constraint on PAE to tackle the above-mentioned challenge. Our model results in higher revenue and better user experience than state-of-the-art baselines in offline experiments. Moreover, our model demonstrates a significant improvement in the online A/B test and has been fully deployed on Meituan feed to serve more than 300 millions of customers.

Jiuxiang Gu, Zhenhua Wang, Jason Kuen et al.

In the last few years, deep learning has led to very good performance on a variety of problems, such as visual recognition, speech recognition and natural language processing. Among different types of deep neural networks, convolutional neural networks have been most extensively studied. Leveraging on the rapid growth in the amount of the annotated data and the great improvements in the strengths of graphics processor units, the research on convolutional neural networks has been emerged swiftly and achieved state-of-the-art results on various tasks. In this paper, we provide a broad survey of the recent advances in convolutional neural networks. We detailize the improvements of CNN on different aspects, including layer design, activation function, loss function, regularization, optimization and fast computation. Besides, we also introduce various applications of convolutional neural networks in computer vision, speech and natural language processing.

Zhenhua Wang, Xingxing Wang, Gang Wang

We propose a novel approach to enhance the discriminability of Convolutional Neural Networks (CNN). The key idea is to build a tree structure that could progressively learn fine-grained features to distinguish a subset of classes, by learning features only among these classes. Such features are expected to be more discriminative, compared to features learned for all the classes. We develop a new algorithm to effectively learn the tree structure from a large number of classes. Experiments on large-scale image classification tasks demonstrate that our method could boost the performance of a given basic CNN model. Our method is quite general, hence it can potentially be used in combination with many other deep learning models.

Zhen Zuo, Bing Shuai, Gang Wang et al.

Existing deep convolutional neural networks (CNNs) have shown their great success on image classification. CNNs mainly consist of convolutional and pooling layers, both of which are performed on local image areas without considering the dependencies among different image regions. However, such dependencies are very important for generating explicit image representation. In contrast, recurrent neural networks (RNNs) are well known for their ability of encoding contextual information among sequential data, and they only require a limited number of network parameters. General RNNs can hardly be directly applied on non-sequential data. Thus, we proposed the hierarchical RNNs (HRNNs). In HRNNs, each RNN layer focuses on modeling spatial dependencies among image regions from the same scale but different locations. While the cross RNN scale connections target on modeling scale dependencies among regions from the same location but different scales. Specifically, we propose two recurrent neural network models: 1) hierarchical simple recurrent network (HSRN), which is fast and has low computational cost; and 2) hierarchical long-short term memory recurrent network (HLSTM), which performs better than HSRN with the price of more computational cost. In this manuscript, we integrate CNNs with HRNNs, and develop end-to-end convolutional hierarchical recurrent neural networks (C-HRNNs). C-HRNNs not only make use of the representation power of CNNs, but also efficiently encodes spatial and scale dependencies among different image regions. On four of the most challenging object/scene image classification benchmarks, our C-HRNNs achieve state-of-the-art results on Places 205, SUN 397, MIT indoor, and competitive results on ILSVRC 2012.

Xiaojiang Peng, Limin Wang, Xingxing Wang et al.

Video based action recognition is one of the important and challenging problems in computer vision research. Bag of Visual Words model (BoVW) with local features has become the most popular method and obtained the state-of-the-art performance on several realistic datasets, such as the HMDB51, UCF50, and UCF101. BoVW is a general pipeline to construct a global representation from a set of local features, which is mainly composed of five steps: (i) feature extraction, (ii) feature pre-processing, (iii) codebook generation, (iv) feature encoding, and (v) pooling and normalization. Many efforts have been made in each step independently in different scenarios and their effect on action recognition is still unknown. Meanwhile, video data exhibits different views of visual pattern, such as static appearance and motion dynamics. Multiple descriptors are usually extracted to represent these different views. Many feature fusion methods have been developed in other areas and their influence on action recognition has never been investigated before. This paper aims to provide a comprehensive study of all steps in BoVW and different fusion methods, and uncover some good practice to produce a state-of-the-art action recognition system. Specifically, we explore two kinds of local features, ten kinds of encoding methods, eight kinds of pooling and normalization strategies, and three kinds of fusion methods. We conclude that every step is crucial for contributing to the final recognition rate. Furthermore, based on our comprehensive study, we propose a simple yet effective representation, called hybrid representation, by exploring the complementarity of different BoVW frameworks and local descriptors. Using this representation, we obtain the state-of-the-art on the three challenging datasets: HMDB51 (61.1%), UCF50 (92.3%), and UCF101 (87.9%).