Searcharxiv⌕ Search

arXiv subjects

Yanhua Cheng

Publications and source records attributed to Yanhua Cheng.

At least 19 recordsLinked to original sources
arXiv2026

TWICE: Two-Clock, Two-Window Learning for Long-Horizon Conversion Prediction in Online Advertising

Long-horizon conversion prediction under delayed feedback creates a two-clock, two-window learning problem in online advertising. A short base observation window releases recent clicks on the click clock before their outcomes mature, whereas conversions continue to arrive on the conversion clock throughout a longer target conversion window. The click clock provides timely but partially observed status supervision. The conversion clock reveals long-tail delays, but the delay composition within an arrival-time slice is weighted by historical click cohorts with different traffic volumes and target-window conversion rates. We present TWICE, a framework that factorizes long-horizon post-click conversion rate (CVR) into a target-window conversion probability and a grouped elapsed-delay cumulative distribution function (CDF). The two clocks provide complementary supervision. Click-clock records train the target-window CVR head through a current-status likelihood over the base observation window. Newly arrived conversions train the delay model on the conversion clock. To account for the cohort mixture, TWICE uses fixed click-time predicted CVR (pCVR) mass as cohort exposure in an arrival-conditioned likelihood. This accounts for differences in cohort traffic and conversion propensity. The resulting aggregate records are self-contained. A single learned CDF produces monotone predictions for all requested horizons up to the target conversion window. Serving requires neither historical lookup nor convolution. Experiments on a public benchmark and an industrial advertising dataset demonstrate the effectiveness of TWICE. In an online A/B test in Kwai's advertising system, TWICE increased expected revenue, revenue, and conversions by 2.486%, 1.858%, and 2.061%, respectively. It was subsequently deployed to full traffic.

cs.LG↗
arXiv2026

ACPO: Asymmetric Credit Policy Optimization via Mode-Local Entropy Surrogate

Outcome-supervised reinforcement learning scales to verifiable reasoning tasks, but trajectory-level rewards assign the same outcome signal to all sampled tokens, overlooking their unequal contributions to the reasoning process. Entropy provides a natural indicator of the model's decision state, yet using it for token-level credit assignment presents two key challenges: long-tail probabilities in large vocabularies corrupt both entropy values and gradients, and uncertainty carries distinct semantics across positive- and non-positive-advantage trajectories. We propose Asymmetric Credit Policy Optimization (ACPO), which replaces global entropy with the complement of the top-token probability as a mode-local proxy. Guided by gradient analysis, ACPO incorporates mismatch routing and saturation correction to shape policy updates into the desired asymmetric form, emphasizing uncertain decisions on positive trajectories while penalizing confident regions on failed ones. Theoretically, ACPO locally preserves the advantage direction while bounding surrogate error. Experiments on mathematical and coding reasoning benchmarks, including AIME 2025 and HumanEval Pro, show that ACPO consistently outperforms both entropy-aware methods (e.g., 80/20, GTPO) and strong outcome-supervised RL baselines (e.g., DAPO, SAPO).

cs.LG↗
arXiv2026

Recommendation as Generation: Unifying Personalized Video Generation and Recommendation at Industrial Scale

Traditional short-video recommendation systems match user interest to a fixed pool of pre-produced videos, which limits their ability to capture fine-grained and dynamic preferences. We propose Recommendation-as-Generation (RaG), a new paradigm that generates personalized videos on demand from inferred user interest. Our framework unifies generative recommendation and video generation through shared semantic IDs (SIDs), which disentangle video representation into content semantics and creative style semantics, enabling both fine-grained modeling of user interest and controllable generation of interest-aligned videos. We further develop Video Generation Agents (VGAs) that are conditioned on inferred SIDs to drive hierarchical planning and refinement for video creation, including visual composition, audio alignment, and artistic effect enhancement. To optimize the framework, we effectively introduce a synergistic cross-domain reward learning mechanism that jointly enforces interest alignment, user feedback, and video quality assessment. We deploy RaG on an industrial-scale platform with over 400 million daily active users and evaluate it in a revenue-critical advertising scenario. Online A/B tests show up to 1.87% ad revenue improvement compared to a strong production GRM baseline, demonstrating its effectiveness in driving further revenue gains beyond generative recommendation. Our results highlight a closed-loop generative system as a promising paradigm for integrating personalized video generation into recommendation.

cs.IR↗
arXiv2026

FBOS-RL: Feedback-Driven Bi-Objective Synergistic Reinforcement Learning

Reinforcement learning has become a cornerstone for aligning and unlocking the reasoning capabilities of large-scale models. At its core, the training loop of GRPO and its variants alternates between rollout sampling and policy update: the policy first samples rollouts from its action space, and then updates its parameters according to the advantages computed over them. Unlike supervised learning, where each gradient step is anchored to an explicit ground-truth target, the optimal gradient direction for updating model parameters in this setting is not known a priori; the high-quality rollouts drawn during the sampling stage therefore act as the implicit "teacher" that guides every parameter update. However, mainstream RL algorithms such as GRPO adopt a simple sampling scheme that conditions all rollouts on the same original prompt. When a task lies beyond the policy model's current capability, this sampling scheme rarely yields a high-quality rollout, leaving the policy model without a meaningful gradient direction when updating its parameters, which causes training to stall. To address this issue, we propose FBOS-RL. Specifically, we let the model perform Feedback-Guided Exploration Enhancement based on the feedback provided by the environment, and on top of this we design two mutually reinforcing training objectives: EPA and ECC. Extensive experiments demonstrate that EPA and ECC can mutually reinforce each other, forming a positive flywheel effect that significantly improves both the training efficiency and the final performance ceiling of reinforcement learning. Specifically, under both an identical number of rollouts and the same number of training steps, FBOS-RL learns substantially faster than GRPO and feedback-based baselines and ultimately attains a higher performance ceiling, while exhibiting higher policy entropy and lower gradient norms throughout training.

cs.LG↗
arXiv2025

VQL: An End-to-End Context-Aware Vector Quantization Attention for Ultra-Long User Behavior Modeling

In large-scale recommender systems, ultra-long user behavior sequences encode rich signals of evolving interests. Extending sequence length generally improves accuracy, but directly modeling such sequences in production is infeasible due to latency and memory constraints. Existing solutions fall into two categories: (1) top-k retrieval, which truncates the sequence and may discard most attention mass when L >> k; and (2) encoder-based compression, which preserves coverage but often over-compresses and fails to incorporate key context such as temporal gaps or target-aware signals. Neither class achieves a good balance of low-loss compression, context awareness, and efficiency. We propose VQL, a context-aware Vector Quantization Attention framework for ultra-long behavior modeling, with three innovations. (1) Key-only quantization: only attention keys are quantized, while values remain intact; we prove that softmax normalization yields an error bound independent of sequence length, and a codebook loss directly supervises quantization quality. This also enables L-free inference via offline caches. (2) Multi-scale quantization: attention heads are partitioned into groups, each with its own small codebook, which reduces quantization error while keeping cache size fixed. (3) Efficient context injection: static features (e.g., item category, modality) are directly integrated, and relative position is modeled via a separable temporal kernel. All context is injected without enlarging the codebook, so cached representations remain query-independent. Experiments on three large-scale datasets (KuaiRand-1K, KuaiRec, TMALL) show that VQL consistently outperforms strong baselines, achieving higher accuracy while reducing inference latency, establishing a new state of the art in balancing accuracy and efficiency for ultra-long sequence recommendation.

cs.IR↗
arXiv2025

Aggregate and Broadcast: Scalable and Efficient Feature Interaction for Recommender Systems

Feature interaction is a core ingredient in ranking models for large-scale recommender systems, yet making it both expressive and efficiently scalable remains challenging. Exhaustive pairwise interaction is powerful but incurs quadratic complexity in the number of tokens/features, while many efficient alternatives rely on restrictive structures that limit information exchange. We further identify two common bottlenecks in practice: (1) early aggregation of behavior sequences compresses fine-grained signals, making it difficult for deeper layers to reuse item-level details; and (2) late fusion injects task signals only at the end, preventing task objectives from directly guiding the interaction process. To address these issues, we propose the Information Flow Network (INFNet), a lightweight architecture that enables scalable, task-aware feature interaction with linear complexity. INFNet represents categorical features, behavior sequences, and task identifiers as tokens, and introduces a small set of hub tokens for each group to serve as communication hubs. Interaction is realized through an efficient aggregate-and-broadcast information flow: hub tokens aggregate global context across groups via cross-attention, and a lightweight gated broadcast unit injects the refined context back to update the categorical, sequence, and task tokens. This design supports width-preserving stacking that preserves item-level signals in sequence and enables task-guided interaction throughout the network, while reducing interaction cost from quadratic to linear in the number of feature tokens. Experiments on a public benchmark and a large-scale industrial dataset demonstrate that INFNet consistently outperforms strong baselines and exhibits strong scaling behavior. In a commercial online advertising system, deploying INFNet improves revenue by +1.587% and click-through rate by +1.155%.

cs.IR↗
arXiv2025

EBaReT: Expert-guided Bag Reward Transformer for Auto Bidding

Reinforcement learning has been widely applied in automated bidding. Traditional approaches model bidding as a Markov Decision Process (MDP). Recently, some studies have explored using generative reinforcement learning methods to address long-term dependency issues in bidding environments. Although effective, these methods typically rely on supervised learning approaches, which are vulnerable to low data quality due to the amount of sub-optimal bids and low probability rewards resulting from the low click and conversion rates. Unfortunately, few studies have addressed these challenges. In this paper, we formalize the automated bidding as a sequence decision-making problem and propose a novel Expert-guided Bag Reward Transformer (EBaReT) to address concerns related to data quality and uncertainty rewards. Specifically, to tackle data quality issues, we generate a set of expert trajectories to serve as supplementary data in the training process and employ a Positive-Unlabeled (PU) learning-based discriminator to identify expert transitions. To ensure the decision also meets the expert level, we further design a novel expert-guided inference strategy. Moreover, to mitigate the uncertainty of rewards, we consider the transitions within a certain period as a "bag" and carefully design a reward function that leads to a smoother acquisition of rewards. Extensive experiments demonstrate that our model achieves superior performance compared to state-of-the-art bidding methods.

cs.LG↗
arXiv2025

Reward Balancing Revisited: Enhancing Offline Reinforcement Learning for Recommender Systems

Offline reinforcement learning (RL) has emerged as a prevalent and effective methodology for real-world recommender systems, enabling learning policies from historical data and capturing user preferences. In offline RL, reward shaping encounters significant challenges, with past efforts to incorporate prior strategies for uncertainty to improve world models or penalize underexplored state-action pairs. Despite these efforts, a critical gap remains: the simultaneous balancing of intrinsic biases in world models and the diversity of policy recommendations. To address this limitation, we present an innovative offline RL framework termed Reallocated Reward for Recommender Systems (R3S). By integrating inherent model uncertainty to tackle the intrinsic fluctuations in reward predictions, we boost diversity for decision-making to align with a more interactive paradigm, incorporating extra penalizers with decay that deter actions leading to diminished state variety at both local and global scales. The experimental results demonstrate that R3S improves the accuracy of world models and efficiently harmonizes the heterogeneous preferences of the users.

cs.IR↗
arXiv2025

Optimal Return-to-Go Guided Decision Transformer for Auto-Bidding in Advertisement

In the realm of online advertising, advertisers partake in ad auctions to obtain advertising slots, frequently taking advantage of auto-bidding tools provided by demand-side platforms. To improve the automation of these bidding systems, we adopt generative models, namely the Decision Transformer (DT), to tackle the difficulties inherent in automated bidding. Applying the Decision Transformer to the auto-bidding task enables a unified approach to sequential modeling, which efficiently overcomes short-sightedness by capturing long-term dependencies between past bidding actions and user behavior. Nevertheless, conventional DT has certain drawbacks: (1) DT necessitates a preset return-to-go (RTG) value before generating actions, which is not inherently produced; (2) The policy learned by DT is restricted by its training data, which is consists of mixed-quality trajectories. To address these challenges, we introduce the R* Decision Transformer (R* DT), developed in a three-step process: (1) R DT: Similar to traditional DT, R DT stores actions based on state and RTG value, as well as memorizing the RTG for a given state using the training set; (2) R^ DT: We forecast the highest value (within the training set) of RTG for a given state, deriving a suboptimal policy based on the current state and the forecasted supreme RTG value; (3) R* DT: Based on R^ DT, we generate trajectories and select those with high rewards (using a simulator) to augment our training dataset. This data enhancement has been shown to improve the RTG of trajectories in the training data and gradually leads the suboptimal policy towards optimality. Comprehensive tests on a publicly available bidding dataset validate the R* DT's efficacy and highlight its superiority when dealing with mixed-quality trajectories.

cs.LG↗
arXiv2025

Learning Monotonic Probabilities with a Generative Cost Model

In many machine learning tasks, it is often necessary for the relationship between input and output variables to be monotonic, including both strictly monotonic and implicitly monotonic relationships. Traditional methods for maintaining monotonicity mainly rely on construction or regularization techniques, whereas this paper shows that the issue of strict monotonic probability can be viewed as a partial order between an observable revenue variable and a latent cost variable. This perspective enables us to reformulate the monotonicity challenge into modeling the latent cost variable. To tackle this, we introduce a generative network for the latent cost variable, termed the Generative Cost Model (GCM), which inherently addresses the strict monotonic problem, and propose the Implicit Generative Cost Model (IGCM) to address the implicit monotonic problem. We further validate our approach with a numerical simulation of quantile regression and conduct multiple experiments on public datasets, showing that our method significantly outperforms existing monotonic modeling techniques. The code for our experiments can be found at https://github.com/tyxaaron/GCM.

cs.LG↗
arXiv2025

BBQRec: Behavior-Bind Quantization for Multi-Modal Sequential Recommendation

Multi-modal sequential recommendation systems leverage auxiliary signals (e.g., text, images) to alleviate data sparsity in user-item interactions. While recent methods exploit large language models to encode modalities into discrete semantic IDs for autoregressive prediction, we identify two critical limitations: (1) Existing approaches adopt fragmented quantization, where modalities are independently mapped to semantic spaces misaligned with behavioral objectives, and (2) Over-reliance on semantic IDs disrupts inter-modal semantic coherence, thereby weakening the expressive power of multi-modal representations for modeling diverse user preferences. To address these challenges, we propose a Behavior-Bind multi-modal Quantization for Sequential Recommendation (BBQRec for short) featuring dual-aligned quantization and semantics-aware sequence modeling. First, our behavior-semantic alignment module disentangles modality-agnostic behavioral patterns from noisy modality-specific features through contrastive codebook learning, ensuring semantic IDs are inherently tied to recommendation tasks. Second, we design a discretized similarity reweighting mechanism that dynamically adjusts self-attention scores using quantized semantic relationships, preserving multi-modal synergies while avoiding invasive modifications to the sequence modeling architecture. Extensive evaluations across four real-world benchmarks demonstrate BBQRec's superiority over the state-of-the-art baselines.

cs.IR↗
arXiv2025

CHIME: A Compressive Framework for Holistic Interest Modeling

Modeling holistic user interests is important for improving recommendation systems but is challenged by high computational cost and difficulty in handling diverse information with full behavior context. Existing search-based methods might lose critical signals during behavior selection. To overcome these limitations, we propose CHIME: A Compressive Framework for Holistic Interest Modeling. It uses adapted large language models to encode complete user behaviors with heterogeneous inputs. We introduce multi-granular contrastive learning objectives to capture both persistent and transient interest patterns and apply residual vector quantization to generate compact embeddings. CHIME demonstrates superior ranking performance across diverse datasets, establishing a robust solution for scalable holistic interest modeling in recommendation systems.

cs.IR↗
arXiv2024

S-Diff: An Anisotropic Diffusion Model for Collaborative Filtering in Spectral Domain

Recovering user preferences from user-item interaction matrices is a key challenge in recommender systems. While diffusion models can sample and reconstruct preferences from latent distributions, they often fail to capture similar users' collective preferences effectively. Additionally, latent variables degrade into pure Gaussian noise during the forward process, lowering the signal-to-noise ratio, which in turn degrades performance. To address this, we propose S-Diff, inspired by graph-based collaborative filtering, better to utilize low-frequency components in the graph spectral domain. S-Diff maps user interaction vectors into the spectral domain and parameterizes diffusion noise to align with graph frequency. This anisotropic diffusion retains significant low-frequency components, preserving a high signal-to-noise ratio. S-Diff further employs a conditional denoising network to encode user interactions, recovering true preferences from noisy data. This method achieves strong results across multiple datasets.

cs.IR↗
arXiv2024

Enhancing Instruction-Following Capability of Visual-Language Models by Reducing Image Redundancy

Large Language Models (LLMs) have strong instruction-following capability to interpret and execute tasks as directed by human commands. Multimodal Large Language Models (MLLMs) have inferior instruction-following ability compared to LLMs. However, there is a significant gap in the instruction-following capabilities between the MLLMs and LLMs. In this study, we conduct a pilot experiment, which demonstrates that spatially down-sampling visual tokens significantly enhances the instruction-following capability of MLLMs. This is attributed to the substantial redundancy in visual modality. However, this intuitive method severely impairs the MLLM's multimodal understanding capability. In this paper, we propose Visual-Modality Token Compression (VMTC) and Cross-Modality Attention Inhibition (CMAI) strategies to alleviate this gap between MLLMs and LLMs by inhibiting the influence of irrelevant visual tokens during content generation, increasing the instruction-following ability of the MLLMs while retaining their multimodal understanding capacity. In VMTC module, the primary tokens are retained and the redundant tokens are condensed by token clustering and merging. In CMAI process, we aggregate text-to-image attentions by text-to-text attentions to obtain a text-to-image focus score. Attention inhibition is performed on the text-image token pairs with low scores. Our comprehensive experiments over instruction-following capabilities and VQA-V2, GQA, TextVQA, MME and MMBench five benchmarks, demonstrate that proposed strategy significantly enhances the instruction following capability of MLLMs while preserving the ability to understand and process multimodal inputs.

cs.CV↗
arXiv2024

Knowledge Condensation and Reasoning for Knowledge-based VQA

Knowledge-based visual question answering (KB-VQA) is a challenging task, which requires the model to leverage external knowledge for comprehending and answering questions grounded in visual content. Recent studies retrieve the knowledge passages from external knowledge bases and then use them to answer questions. However, these retrieved knowledge passages often contain irrelevant or noisy information, which limits the performance of the model. To address the challenge, we propose two synergistic models: Knowledge Condensation model and Knowledge Reasoning model. We condense the retrieved knowledge passages from two perspectives. First, we leverage the multimodal perception and reasoning ability of the visual-language models to distill concise knowledge concepts from retrieved lengthy passages, ensuring relevance to both the visual content and the question. Second, we leverage the text comprehension ability of the large language models to summarize and condense the passages into the knowledge essence which helps answer the question. These two types of condensed knowledge are then seamlessly integrated into our Knowledge Reasoning model, which judiciously navigates through the amalgamated information to arrive at the conclusive answer. Extensive experiments validate the superiority of the proposed method. Compared to previous methods, our method achieves state-of-the-art performance on knowledge-based VQA datasets (65.1% on OK-VQA and 60.1% on A-OKVQA) without resorting to the knowledge produced by GPT-3 (175B).

cs.CV↗
arXiv2024

Towards Efficient and Effective Text-to-Video Retrieval with Coarse-to-Fine Visual Representation Learning

In recent years, text-to-video retrieval methods based on CLIP have experienced rapid development. The primary direction of evolution is to exploit the much wider gamut of visual and textual cues to achieve alignment. Concretely, those methods with impressive performance often design a heavy fusion block for sentence (words)-video (frames) interaction, regardless of the prohibitive computation complexity. Nevertheless, these approaches are not optimal in terms of feature utilization and retrieval efficiency. To address this issue, we adopt multi-granularity visual feature learning, ensuring the model's comprehensiveness in capturing visual content features spanning from abstract to detailed levels during the training phase. To better leverage the multi-granularity features, we devise a two-stage retrieval architecture in the retrieval phase. This solution ingeniously balances the coarse and fine granularity of retrieval content. Moreover, it also strikes a harmonious equilibrium between retrieval effectiveness and efficiency. Specifically, in training phase, we design a parameter-free text-gated interaction block (TIB) for fine-grained video representation learning and embed an extra Pearson Constraint to optimize cross-modal representation learning. In retrieval phase, we use coarse-grained video representations for fast recall of top-k candidates, which are then reranked by fine-grained video representations. Extensive experiments on four benchmarks demonstrate the efficiency and effectiveness. Notably, our method achieves comparable performance with the current state-of-the-art methods while being nearly 50 times faster.

cs.CV↗
arXiv2023

Cross-Domain Product Representation Learning for Rich-Content E-Commerce

The proliferation of short video and live-streaming platforms has revolutionized how consumers engage in online shopping. Instead of browsing product pages, consumers are now turning to rich-content e-commerce, where they can purchase products through dynamic and interactive media like short videos and live streams. This emerging form of online shopping has introduced technical challenges, as products may be presented differently across various media domains. Therefore, a unified product representation is essential for achieving cross-domain product recognition to ensure an optimal user search experience and effective product recommendations. Despite the urgent industrial need for a unified cross-domain product representation, previous studies have predominantly focused only on product pages without taking into account short videos and live streams. To fill the gap in the rich-content e-commerce area, in this paper, we introduce a large-scale cRoss-dOmain Product Ecognition dataset, called ROPE. ROPE covers a wide range of product categories and contains over 180,000 products, corresponding to millions of short videos and live streams. It is the first dataset to cover product pages, short videos, and live streams simultaneously, providing the basis for establishing a unified product representation across different media domains. Furthermore, we propose a Cross-dOmain Product rEpresentation framework, namely COPE, which unifies product representations in different domains through multimodal learning including text and vision. Extensive experiments on downstream tasks demonstrate the effectiveness of COPE in learning a joint feature space for all product domains.

cs.CV↗
arXiv2023

Cross-view Semantic Alignment for Livestreaming Product Recognition

Live commerce is the act of selling products online through live streaming. The customer's diverse demands for online products introduce more challenges to Livestreaming Product Recognition. Previous works have primarily focused on fashion clothing data or utilize single-modal input, which does not reflect the real-world scenario where multimodal data from various categories are present. In this paper, we present LPR4M, a large-scale multimodal dataset that covers 34 categories, comprises 3 modalities (image, video, and text), and is 50x larger than the largest publicly available dataset. LPR4M contains diverse videos and noise modality pairs while exhibiting a long-tailed distribution, resembling real-world problems. Moreover, a cRoss-vIew semantiC alignmEnt (RICE) model is proposed to learn discriminative instance features from the image and video views of the products. This is achieved through instance-level contrastive learning and cross-view patch-level feature propagation. A novel Patch Feature Reconstruction loss is proposed to penalize the semantic misalignment between cross-view patches. Extensive experiments demonstrate the effectiveness of RICE and provide insights into the importance of dataset diversity and expressivity. The dataset and code are available at https://github.com/adxcreative/RICE

cs.CV↗