arXiv subjects

Anxiang Zeng

Publications and source records attributed to Anxiang Zeng.

At least 37 records · Page 2Linked to original sources

arXiv2025

MUG-V 10B: High-efficiency Training Pipeline for Large Video Generation Models

In recent years, large-scale generative models for visual content (\textit{e.g.,} images, videos, and 3D objects/scenes) have made remarkable progress. However, training large-scale video generation models remains particularly challenging and resource-intensive due to cross-modal text-video alignment, the long sequences involved, and the complex spatiotemporal dependencies. To address these challenges, we present a training framework that optimizes four pillars: (i) data processing, (ii) model architecture, (iii) training strategy, and (iv) infrastructure for large-scale video generation models. These optimizations delivered significant efficiency gains and performance improvements across all stages of data preprocessing, video compression, parameter scaling, curriculum-based pretraining, and alignment-focused post-training. Our resulting model, MUG-V 10B, matches recent state-of-the-art video generators overall and, on e-commerce-oriented video generation tasks, surpasses leading open-source baselines in human evaluations. More importantly, we open-source the complete stack, including model weights, Megatron-Core-based large-scale training code, and inference pipelines for video generation and enhancement. To our knowledge, this is the first public release of large-scale video generation training code that exploits Megatron-Core to achieve high training efficiency and near-linear multi-node scaling, details are available in https://github.com/Shopee-MUG/MUG-V.

cs.CV↗

arXiv2025

Mid-Training of Large Language Models: A Survey

Large language models (LLMs) are typically developed through large-scale pre-training followed by task-specific fine-tuning. Recent advances highlight the importance of an intermediate mid-training stage, where models undergo multiple annealing-style phases that refine data quality, adapt optimization schedules, and extend context length. This stage mitigates diminishing returns from noisy tokens, stabilizes convergence, and expands model capability in late training. Its effectiveness can be explained through gradient noise scale, the information bottleneck, and curriculum learning, which together promote generalization and abstraction. Despite widespread use in state-of-the-art systems, there has been no prior survey of mid-training as a unified paradigm. We introduce the first taxonomy of LLM mid-training spanning data distribution, learning-rate scheduling, and long-context extension. We distill practical insights, compile evaluation benchmarks, and report gains to enable structured comparisons across models. We also identify open challenges and propose avenues for future research and practice.

cs.CL↗

arXiv2025

OnePiece: Bringing Context Engineering and Reasoning to Industrial Cascade Ranking System

Despite the growing interest in replicating the scaled success of large language models (LLMs) in industrial search and recommender systems, most existing industrial efforts remain limited to transplanting Transformer architectures, which bring only incremental improvements over strong Deep Learning Recommendation Models (DLRMs). From a first principle perspective, the breakthroughs of LLMs stem not only from their architectures but also from two complementary mechanisms: context engineering, which enriches raw input queries with contextual cues to better elicit model capabilities, and multi-step reasoning, which iteratively refines model outputs through intermediate reasoning paths. However, these two mechanisms and their potential to unlock substantial improvements remain largely underexplored in industrial ranking systems. In this paper, we propose OnePiece, a unified framework that seamlessly integrates LLM-style context engineering and reasoning into both retrieval and ranking models of industrial cascaded pipelines. OnePiece is built on a pure Transformer backbone and further introduces three key innovations: (1) structured context engineering, which augments interaction history with preference and scenario signals and unifies them into a structured tokenized input sequence for both retrieval and ranking; (2) block-wise latent reasoning, which equips the model with multi-step refinement of representations and scales reasoning bandwidth via block size; (3) progressive multi-task training, which leverages user feedback chains to effectively supervise reasoning steps during training. OnePiece has been deployed in the main personalized search scenario of Shopee and achieves consistent online gains across different key business metrics, including over $+2\%$ GMV/UU and a $+2.90\%$ increase in advertising revenue.

cs.IR↗

arXiv2025

Compass-Thinker-7B Technical Report

Recent R1-Zero-like research further demonstrates that reasoning extension has given large language models (LLMs) unprecedented reasoning capabilities, and Reinforcement Learning is the core technology to elicit its complex reasoning. However, conducting RL experiments directly on hyperscale models involves high computational costs and resource demands, posing significant risks. We propose the Compass-Thinker-7B model, which aims to explore the potential of Reinforcement Learning with less computational resources and costs, and provides insights for further research into RL recipes for larger models. Compass-Thinker-7B is trained from an open source model through a specially designed Reinforcement Learning Pipeline. We curate a dataset of 30k verifiable mathematics problems for the Reinforcement Learning Pipeline. By configuring data and training settings with different difficulty distributions for different stages, the potential of the model is gradually released and the training efficiency is improved. Extensive evaluations show that Compass-Thinker-7B possesses exceptional reasoning potential, and achieves superior performance on mathematics compared to the same-sized RL model. Especially in the challenging AIME2024 evaluation, Compass-Thinker-7B achieves 40% accuracy.

cs.AI↗

arXiv2025

Maximum Inner Product is Query-Scaled Nearest Neighbor

Maximum Inner Product Search (MIPS) for high-dimensional vectors is pivotal across databases, information retrieval, and artificial intelligence. Existing methods either reduce MIPS to Nearest Neighbor Search (NNS) while suffering from harmful vector space transformations, or attempt to tackle MIPS directly but struggle to mitigate redundant computations due to the absence of the triangle inequality. This paper presents a novel theoretical framework that equates MIPS with NNS without requiring space transformation, thereby allowing us to leverage advanced graph-based indices for NNS and efficient edge pruning strategies, significantly reducing unnecessary computations. Despite a strong baseline set by our theoretical analysis, we identify and address two persistent challenges to further refine our method: the introduction of the Proximity Graph with Spherical Pathway (PSP), designed to mitigate the issue of MIPS solutions clustering around large-norm vectors, and the implementation of Adaptive Early Termination (AET), which efficiently curtails the excessive exploration once an accuracy bottleneck is reached. Extensive experiments reveal the superiority of our method over existing state-of-the-art techniques in search efficiency, scalability, and practical applicability. Compared with state-of-the-art graph based methods, it achieves an average 35% speed-up in query processing and a 3x reduction in index size. Notably, our approach has been validated and deployed in the search engines of Shopee, a well-known online shopping platform. Our code and an industrial-scale dataset for offline evaluation will also be released to address the absence of e-commerce data in public benchmarks.

cs.DB↗

arXiv2025

Stitching Inner Product and Euclidean Metrics for Topology-aware Maximum Inner Product Search

Maximum Inner Product Search (MIPS) is a fundamental challenge in machine learning and information retrieval, particularly in high-dimensional data applications. Existing approaches to MIPS either rely solely on Inner Product (IP) similarity, which faces issues with local optima and redundant computations, or reduce the MIPS problem to the Nearest Neighbor Search under the Euclidean metric via space projection, leading to topology destruction and information loss. Despite the divergence of the two paradigms, we argue that there is no inherent binary opposition between IP and Euclidean metrics. By stitching IP and Euclidean in the design of indexing and search algorithms, we can significantly enhance MIPS performance. Specifically, this paper explores the theoretical and empirical connections between these two metrics from the MIPS perspective. Our investigation, grounded in graph-based search, reveals that different indexing and search strategies offer distinct advantages for MIPS, depending on the underlying data topology. Building on these insights, we introduce a novel graph-based index called Metric-Amphibious Graph (MAG) and a corresponding search algorithm, Adaptive Navigation with Metric Switch (ANMS). To facilitate parameter tuning for optimal performance, we identify three statistical indicators that capture essential data topology properties and correlate strongly with parameter tuning. Extensive experiments on 12 real-world datasets demonstrate that MAG outperforms existing state-of-the-art methods, achieving up to 4x search speedup while maintaining adaptability and scalability.

cs.DB↗

arXiv2025

RDPO: Real Data Preference Optimization for Physics Consistency Video Generation

Video generation techniques have achieved remarkable advancements in visual quality, yet faithfully reproducing real-world physics remains elusive. Preference-based model post-training may improve physical consistency, but requires costly human-annotated datasets or reward models that are not yet feasible. To address these challenges, we present Real Data Preference Optimisation (RDPO), an annotation-free framework that distills physical priors directly from real-world videos. Specifically, the proposed RDPO reverse-samples real video sequences with a pre-trained generator to automatically build preference pairs that are statistically distinguishable in terms of physical correctness. A multi-stage iterative training schedule then guides the generator to obey physical laws increasingly well. Benefiting from the dynamic information explored from real videos, our proposed RDPO significantly improves the action coherence and physical realism of the generated videos. Evaluations on multiple benchmarks and human evaluations have demonstrated that RDPO achieves improvements across multiple dimensions. The source code and demonstration of this paper are available at: https://wwenxu.github.io/RDPO/

cs.CV↗

arXiv2025

Embed Progressive Implicit Preference in Unified Space for Deep Collaborative Filtering

Embedding-based collaborative filtering, often coupled with nearest neighbor search, is widely deployed in large-scale recommender systems for personalized content selection. Modern systems leverage multiple implicit feedback signals (e.g., clicks, add to cart, purchases) to model user preferences comprehensively. However, prevailing approaches adopt a feedback-wise modeling paradigm, which (1) fails to capture the structured progression of user engagement entailed among different feedback and (2) embeds feedback-specific information into disjoint spaces, making representations incommensurable, increasing system complexity, and leading to suboptimal retrieval performance. A promising alternative is Ordinal Logistic Regression (OLR), which explicitly models discrete ordered relations. However, existing OLR-based recommendation models mainly focus on explicit feedback (e.g., movie ratings) and struggle with implicit, correlated feedback, where ordering is vague and non-linear. Moreover, standard OLR lacks flexibility in handling feedback-dependent covariates, resulting in suboptimal performance in real-world systems. To address these limitations, we propose Generalized Neural Ordinal Logistic Regression (GNOLR), which encodes multiple feature-feedback dependencies into a unified, structured embedding space and enforces feedback-specific dependency learning through a nested optimization framework. Thus, GNOLR enhances predictive accuracy, captures the progression of user engagement, and simplifies the retrieval process. We establish a theoretical comparison with existing paradigms, demonstrating how GNOLR avoids disjoint spaces while maintaining effectiveness. Extensive experiments on ten real-world datasets show that GNOLR significantly outperforms state-of-the-art methods in efficiency and adaptability.

cs.IR↗

arXiv2025

Optimal Transport-Based Token Weighting scheme for Enhanced Preference Optimization

Direct Preference Optimization (DPO) has emerged as a promising framework for aligning Large Language Models (LLMs) with human preferences by directly optimizing the log-likelihood difference between chosen and rejected responses. However, existing methods assign equal importance to all tokens in the response, while humans focus on more meaningful parts. This leads to suboptimal preference optimization, as irrelevant or noisy tokens disproportionately influence DPO loss. To address this limitation, we propose \textbf{O}ptimal \textbf{T}ransport-based token weighting scheme for enhancing direct \textbf{P}reference \textbf{O}ptimization (OTPO). By emphasizing semantically meaningful token pairs and de-emphasizing less relevant ones, our method introduces a context-aware token weighting scheme that yields a more contrastive reward difference estimate. This adaptive weighting enhances reward stability, improves interpretability, and ensures that preference optimization focuses on meaningful differences between responses. Extensive experiments have validated OTPO's effectiveness in improving instruction-following ability across various settings\footnote{Code is available at https://github.com/Mimasss2/OTPO.}.

cs.CL↗

arXiv2025

Establishing Reliability Metrics for Reward Models in Large Language Models

The reward model (RM) that represents human preferences plays a crucial role in optimizing the outputs of large language models (LLMs), e.g., through reinforcement learning from human feedback (RLHF) or rejection sampling. However, a long challenge for RM is its uncertain reliability, i.e., LLM outputs with higher rewards may not align with actual human preferences. Currently, there is a lack of a convincing metric to quantify the reliability of RMs. To bridge this gap, we propose the \textit{\underline{R}eliable at \underline{$η$}} (RETA) metric, which directly measures the reliability of an RM by evaluating the average quality (scored by an oracle) of the top $η$ quantile responses assessed by an RM. On top of RETA, we present an integrated benchmarking pipeline that allows anyone to evaluate their own RM without incurring additional Oracle labeling costs. Extensive experimental studies demonstrate the superior stability of RETA metric, providing solid evaluations of the reliability of various publicly available and proprietary RMs. When dealing with an unreliable RM, we can use the RETA metric to identify the optimal quantile from which to select the responses.

cs.AI↗

arXiv2024

Residual Multi-Task Learner for Applied Ranking

Modern e-commerce platforms rely heavily on modeling diverse user feedback to provide personalized services. Consequently, multi-task learning has become an integral part of their ranking systems. However, existing multi-task learning methods encounter two main challenges: some lack explicit modeling of task relationships, resulting in inferior performance, while others have limited applicability due to being computationally intensive, having scalability issues, or relying on strong assumptions. To address these limitations and better fit our real-world scenario, pre-rank in Shopee Search, we introduce in this paper ResFlow, a lightweight multi-task learning framework that enables efficient cross-task information sharing via residual connections between corresponding layers of task networks. Extensive experiments on datasets from various scenarios and modalities demonstrate its superior performance and adaptability over state-of-the-art methods. The online A/B tests in Shopee Search showcase its practical value in large-scale industrial applications, evidenced by a 1.29% increase in OPU (order-per-user) without additional system latency. ResFlow is now fully deployed in the pre-rank module of Shopee Search. To facilitate efficient online deployment, we propose a novel offline metric Weighted Recall@K, which aligns well with our online metric OPU, addressing the longstanding online-offline metric misalignment issue. Besides, we propose to fuse scores from the multiple tasks additively when ranking items, which outperforms traditional multiplicative fusion. The code is released at https://github.com/BrunoTruthAlliance/ResFlow

cs.IR↗

arXiv2024

Transformer-empowered Multi-modal Item Embedding for Enhanced Image Search in E-Commerce

Over the past decade, significant advances have been made in the field of image search for e-commerce applications. Traditional image-to-image retrieval models, which focus solely on image details such as texture, tend to overlook useful semantic information contained within the images. As a result, the retrieved products might possess similar image details, but fail to fulfil the user's search goals. Moreover, the use of image-to-image retrieval models for products containing multiple images results in significant online product feature storage overhead and complex mapping implementations. In this paper, we report the design and deployment of the proposed Multi-modal Item Embedding Model (MIEM) to address these limitations. It is capable of utilizing both textual information and multiple images about a product to construct meaningful product features. By leveraging semantic information from images, MIEM effectively supplements the image search process, improving the overall accuracy of retrieval results. MIEM has become an integral part of the Shopee image search platform. Since its deployment in March 2023, it has achieved a remarkable 9.90% increase in terms of clicks per user and a 4.23% boost in terms of orders per user for the image search feature on the Shopee e-commerce platform.

cs.CV↗

arXiv2023

Recurrent Temporal Revision Graph Networks

Temporal graphs offer more accurate modeling of many real-world scenarios than static graphs. However, neighbor aggregation, a critical building block of graph networks, for temporal graphs, is currently straightforwardly extended from that of static graphs. It can be computationally expensive when involving all historical neighbors during such aggregation. In practice, typically only a subset of the most recent neighbors are involved. However, such subsampling leads to incomplete and biased neighbor information. To address this limitation, we propose a novel framework for temporal neighbor aggregation that uses the recurrent neural network with node-wise hidden states to integrate information from all historical neighbors for each node to acquire the complete neighbor information. We demonstrate the superior theoretical expressiveness of the proposed framework as well as its state-of-the-art performance in real-world applications. Notably, it achieves a significant +9.6% improvement on averaged precision in a real-world Ecommerce dataset over existing methods on 2-layer models.

cs.LG↗

arXiv2023

Clustered Embedding Learning for Recommender Systems

In recent years, recommender systems have advanced rapidly, where embedding learning for users and items plays a critical role. A standard method learns a unique embedding vector for each user and item. However, such a method has two important limitations in real-world applications: 1) it is hard to learn embeddings that generalize well for users and items with rare interactions on their own; and 2) it may incur unbearably high memory costs when the number of users and items scales up. Existing approaches either can only address one of the limitations or have flawed overall performances. In this paper, we propose Clustered Embedding Learning (CEL) as an integrated solution to these two problems. CEL is a plug-and-play embedding learning framework that can be combined with any differentiable feature interaction model. It is capable of achieving improved performance, especially for cold users and items, with reduced memory cost. CEL enables automatic and dynamic clustering of users and items in a top-down fashion, where clustered entities jointly learn a shared embedding. The accelerated version of CEL has an optimal time complexity, which supports efficient online updates. Theoretically, we prove the identifiability and the existence of a unique optimal number of clusters for CEL in the context of nonnegative matrix factorization. Empirically, we validate the effectiveness of CEL on three public datasets and one business dataset, showing its consistently superior performance against current state-of-the-art methods. In particular, when incorporating CEL into the business model, it brings an improvement of $+0.6\%$ in AUC, which translates into a significant revenue gain; meanwhile, the size of the embedding table gets $2650$ times smaller.

cs.AI↗

arXiv2021

Learning an Adaptive Meta Model-Generator for Incrementally Updating Recommender Systems

Recommender Systems (RSs) in real-world applications often deal with billions of user interactions daily. To capture the most recent trends effectively, it is common to update the model incrementally using only the newly arrived data. However, this may impede the model's ability to retain long-term information due to the potential overfitting and forgetting issues. To address this problem, we propose a novel Adaptive Sequential Model Generation (ASMG) framework, which generates a better serving model from a sequence of historical models via a meta generator. For the design of the meta generator, we propose to employ Gated Recurrent Units (GRUs) to leverage its ability to capture the long-term dependencies. We further introduce some novel strategies to apply together with the GRU meta generator, which not only improve its computational efficiency but also enable more accurate sequential modeling. By instantiating the model-agnostic framework on a general deep learning-based RS model, we demonstrate that our method achieves state-of-the-art performance on three public datasets and one industrial dataset.

cs.IR↗

arXiv2021

Diversity Regularized Interests Modeling for Recommender Systems

With the rapid development of E-commerce and the increase in the quantity of items, users are presented with more items hence their interests broaden. It is increasingly difficult to model user intentions with traditional methods, which model the user's preference for an item by combining a single user vector and an item vector. Recently, some methods are proposed to generate multiple user interest vectors and achieve better performance compared to traditional methods. However, empirical studies demonstrate that vectors generated from these multi-interests methods are sometimes homogeneous, which may lead to sub-optimal performance. In this paper, we propose a novel method of Diversity Regularized Interests Modeling (DRIM) for Recommender Systems. We apply a capsule network in a multi-interest extractor to generate multiple user interest vectors. Each interest of the user should have a certain degree of distinction, thus we introduce three strategies as the diversity regularized separator to separate multiple user interest vectors. Experimental results on public and industrial data sets demonstrate the ability of the model to capture different interests of a user and the superior performance of the proposed approach.

cs.IR↗

arXiv2020

Hybrid Interest Modeling for Long-tailed Users

User behavior modeling is a key technique for recommender systems. However, most methods focus on head users with large-scale interactions and hence suffer from data sparsity issues. Several solutions integrate side information such as demographic features and product reviews, another is to transfer knowledge from other rich data sources. We argue that current methods are limited by the strict privacy policy and have low scalability in real-world applications and few works consider the behavioral characteristics behind long-tailed users. In this work, we propose the Hybrid Interest Modeling (HIM) network to hybrid both personalized interest and semi-personalized interest in learning long-tailed users' preferences in the recommendation. To achieve this, we first design the User Behavior Pyramid (UBP) module to capture the fine-grained personalized interest of high confidence from sparse even noisy positive feedbacks. Moreover, the individual interaction is too sparse and not enough for modeling user interest adequately, we design the User Behavior Clustering (UBC) module to learn latent user interest groups with self-supervised learning mechanism novelly, which capture coarse-grained semi-personalized interest from group-item interaction data. Extensive experiments on both public and industrial datasets verify the superiority of HIM compared with the state-of-the-art baselines.

cs.IR↗

arXiv2020

Delayed Feedback Modeling for the Entire Space Conversion Rate Prediction

Estimating post-click conversion rate (CVR) accurately is crucial in E-commerce. However, CVR prediction usually suffers from three major challenges in practice: i) data sparsity: compared with impressions, conversion samples are often extremely scarce; ii) sample selection bias: conventional CVR models are trained with clicked impressions while making inference on the entire space of all impressions; iii) delayed feedback: many conversions can only be observed after a relatively long and random delay since clicks happened, resulting in many false negative labels during training. Previous studies mainly focus on one or two issues while ignoring the others. In this paper, we propose a novel neural network framework ESDF to tackle the above three challenges simultaneously. Unlike existing methods, ESDF models the CVR prediction from a perspective of entire space, and combines the advantage of user sequential behavior pattern and the time delay factor. Specifically, ESDF utilizes sequential behavior of user actions on the entire space with all impressions to alleviate the sample selection bias problem. By sharing the embedding parameters between CTR and CVR networks, data sparsity problem is greatly relieved. Different from conventional delayed feedback methods, ESDF does not make any special assumption about the delay distribution. We discretize the delay time by day slot and model the probability based on survival analysis with deep neural network, which is more practical and suitable for industrial situations. Extensive experiments are conducted to evaluate the effectiveness of our method. To the best of our knowledge, ESDF is the first attempt to unitedly solve the above three challenges in CVR prediction area.

cs.LG↗