arXiv subjects

Yu Zhao

Publications and source records attributed to Yu Zhao.

At least 91 records · Page 5Linked to original sources

arXiv2025

Mobile Gamer Lifetime Value Prediction via Objective Decomposition and Reconstruction

For Internet platforms operating real-time bidding (RTB) advertising service, a comprehensive understanding of user lifetime value (LTV) plays a pivotal role in optimizing advertisement allocation efficiency and maximizing the return on investment (ROI) for advertisement sponsors, thereby facilitating growth of commercialization revenue for the platform. However, the inherent complexity of user LTV distributions induces significant challenges in accurate LTV prediction. Existing state-of-the-art works, which primarily focus on directly learning the LTV distributions through well-designed loss functions, achieve limited success due to their vulnerability to outliers. In this paper, we proposed a novel LTV prediction method to address distribution challenges through an objective decomposition and reconstruction framework. Briefly speaking, based on the in-app purchase characteristics of mobile gamers, our model was designed to first predict the number of transactions at specific prices and then calculate the total payment amount from these intermediate predictions. Our proposed model was evaluated through experiments on real-world industrial dataset, and deployed on the TapTap RTB advertising system for online A/B testing along with the state-of-the-art ZILN model.

cs.IR↗

arXiv2025

MMLongBench: Benchmarking Long-Context Vision-Language Models Effectively and Thoroughly

The rapid extension of context windows in large vision-language models has given rise to long-context vision-language models (LCVLMs), which are capable of handling hundreds of images with interleaved text tokens in a single forward pass. In this work, we introduce MMLongBench, the first benchmark covering a diverse set of long-context vision-language tasks, to evaluate LCVLMs effectively and thoroughly. MMLongBench is composed of 13,331 examples spanning five different categories of downstream tasks, such as Visual RAG and Many-Shot ICL. It also provides broad coverage of image types, including various natural and synthetic images. To assess the robustness of the models to different input lengths, all examples are delivered at five standardized input lengths (8K-128K tokens) via a cross-modal tokenization scheme that combines vision patches and text tokens. Through a thorough benchmarking of 46 closed-source and open-source LCVLMs, we provide a comprehensive analysis of the current models' vision-language long-context ability. Our results show that: i) performance on a single task is a weak proxy for overall long-context capability; ii) both closed-source and open-source models face challenges in long-context vision-language tasks, indicating substantial room for future improvement; iii) models with stronger reasoning ability tend to exhibit better long-context performance. By offering wide task coverage, various image types, and rigorous length control, MMLongBench provides the missing foundation for diagnosing and advancing the next generation of LCVLMs.

cs.CV↗

arXiv2025

T2R-bench: A Benchmark for Generating Article-Level Reports from Real World Industrial Tables

Extensive research has been conducted to explore the capabilities of large language models (LLMs) in table reasoning. However, the essential task of transforming tables information into reports remains a significant challenge for industrial applications. This task is plagued by two critical issues: 1) the complexity and diversity of tables lead to suboptimal reasoning outcomes; and 2) existing table benchmarks lack the capacity to adequately assess the practical application of this task. To fill this gap, we propose the table-to-report task and construct a bilingual benchmark named T2R-bench, where the key information flow from the tables to the reports for this task. The benchmark comprises 457 industrial tables, all derived from real-world scenarios and encompassing 19 industry domains as well as 4 types of industrial tables. Furthermore, we propose an evaluation criteria to fairly measure the quality of report generation. The experiments on 25 widely-used LLMs reveal that even state-of-the-art models like Deepseek-R1 only achieves performance with 62.71 overall score, indicating that LLMs still have room for improvement on T2R-bench.

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

Towards Comprehensive Information-theoretic Multi-view Learning

Information theory has inspired numerous advancements in multi-view learning. Most multi-view methods incorporating information-theoretic principles rely an assumption called multi-view redundancy which states that common information between views is necessary and sufficient for down-stream tasks. This assumption emphasizes the importance of common information for prediction, but inherently ignores the potential of unique information in each view that could be predictive to the task. In this paper, we propose a comprehensive information-theoretic multi-view learning framework named CIML, which discards the assumption of multi-view redundancy. Specifically, CIML considers the potential predictive capabilities of both common and unique information based on information theory. First, the common representation learning maximizes Gacs-Korner common information to extract shared features and then compresses this information to learn task-relevant representations based on the Information Bottleneck (IB). For unique representation learning, IB is employed to achieve the most compressed unique representation for each view while simultaneously minimizing the mutual information between unique and common representations, as well as among different unique representations. Importantly, we theoretically prove that the learned joint representation is predictively sufficient for the downstream task. Extensive experimental results have demonstrated the superiority of our model over several state-of-art methods. The code is released on CIML.

cs.LG↗

arXiv2025

Synergistic Dual Spatial-aware Generation of Image-to-Text and Text-to-Image

In the visual spatial understanding (VSU) area, spatial image-to-text (SI2T) and spatial text-to-image (ST2I) are two fundamental tasks that appear in dual form. Existing methods for standalone SI2T or ST2I perform imperfectly in spatial understanding, due to the difficulty of 3D-wise spatial feature modeling. In this work, we consider modeling the SI2T and ST2I together under a dual learning framework. During the dual framework, we then propose to represent the 3D spatial scene features with a novel 3D scene graph (3DSG) representation that can be shared and beneficial to both tasks. Further, inspired by the intuition that the easier 3D$\to$image and 3D$\to$text processes also exist symmetrically in the ST2I and SI2T, respectively, we propose the Spatial Dual Discrete Diffusion (SD$^3$) framework, which utilizes the intermediate features of the 3D$\to$X processes to guide the hard X$\to$3D processes, such that the overall ST2I and SI2T will benefit each other. On the visual spatial understanding dataset VSD, our system outperforms the mainstream T2I and I2T methods significantly. Further in-depth analysis reveals how our dual learning strategy advances.

cs.CV↗

arXiv2025

TableZoomer: A Collaborative Agent Framework for Large-scale Table Question Answering

While large language models (LLMs) have shown promise in the table question answering (TQA) task through prompt engineering, they face challenges in industrial applications, including structural heterogeneity, difficulties in target data localization, and bottlenecks in complex reasoning. To address these limitations, this paper presents TableZoomer, a novel LLM-powered, programming-based agent framework. It introduces three key innovations: (1) replacing the original fully verbalized table with structured table schema to bridge the semantic gap and reduce computational complexity; (2) a query-aware table zooming mechanism that dynamically generates sub-table schema through column selection and entity linking, significantly improving target localization efficiency; and (3) a Program-of-Thoughts (PoT) strategy that transforms queries into executable code to mitigate numerical hallucination. Additionally, we integrate the reasoning workflow with the ReAct paradigm to enable iterative reasoning. Extensive experiments demonstrate that our framework maintains the usability advantages while substantially enhancing performance and scalability across tables of varying scales. When implemented with the Qwen3-8B-Instruct LLM, TableZoomer achieves accuracy improvements of 19.34% and 25% over conventional PoT methods on the large-scale DataBench dataset and the small-scale Fact Checking task of TableBench dataset, respectively.

cs.CL↗

arXiv2025

A re-identification of six Candidate Gravitationally Lensed Gamma-Ray Bursts

The gravitational lensing effect of gamma-ray bursts (GRBs) holds significant and diverse applications in the field of astronomy. Nevertheless, the identification of millilensing events in GRBs presents substantial challenges. We re-evaluate the gravitational lensing candidacy of six previously proposed GRBs (GRB 081122A, GRB 081126A, GRB 090717A, GRB 110517B, GRB 200716C, and GRB 210812A) using a comprehensive set of temporal and spectral diagnostics. These include $χ^2$ light-curve similarity tests, photon-count-based hardness ratio ($HR_{count}$) comparisons, $T_{90}$ duration measurements, spectral lag, Norris pulse-shape fitting, and both time-resolved and time-integrated spectral analyses. We propose an evaluation framework, any single test that reveals a statistically significant inconsistency between the two pulses is sufficient to reject the lensing hypothesis for that burst.Although certain diagnostics, such as $T_{90}$ and parametric model fits, have known limitations, they are applied and interpreted in conjunction with the more robust, model-independent $χ^2$ and $HR_{count}$ tests. For all six GRBs, at least one diagnostic shows a significant discrepancy, leading us to conclude that none are consistent with a gravitational lensing interpretation.

astro-ph.HE↗

arXiv2025

Constraining interacting dark energy models with the halo concentration-mass relation

The interacting dark energy (IDE) model is a promising alternative cosmological model which has the potential to solve the fine-tuning and coincidence problems by considering the interaction between dark matter and dark energy. Previous studies have shown that the energy exchange between the dark sectors in this model can significantly affect the dark matter halo properties. In this study, utilising a large set of cosmological $N$-body simulations, we analyse the redshift evolution of the halo concentration-mass ($c$-$M$) relation in the IDE model, and show that the $c$-$M$ relation is a sensitive proxy of the interaction strength parameter $ξ_2$, especially at lower redshifts. Furthermore, we construct parametrized formulae to quantify the dependence of the $c$-$M$ relation on $ξ_2$ at redshifts ranging from $z=0$ to $0.6$. Our parametrized formulae provide a useful tool in constraining $ξ_2$ with the observational $c$-$M$ relation. As a first attempt, we use the data from X-ray, gravitational lensing, and galaxy rotational curve observations and obtain a tight constraint on $ξ_2$, i.e. $ξ_2 = 0.071 \pm 0.034$. Our work demonstrates that the halo $c$-$M$ relation, which reflects the halo assembly history, is a powerful probe to constrain the IDE model.

astro-ph.CO↗

arXiv2025

MDEval: Evaluating and Enhancing Markdown Awareness in Large Language Models

Large language models (LLMs) are expected to offer structured Markdown responses for the sake of readability in web chatbots (e.g., ChatGPT). Although there are a myriad of metrics to evaluate LLMs, they fail to evaluate the readability from the view of output content structure. To this end, we focus on an overlooked yet important metric -- Markdown Awareness, which directly impacts the readability and structure of the content generated by these language models. In this paper, we introduce MDEval, a comprehensive benchmark to assess Markdown Awareness for LLMs, by constructing a dataset with 20K instances covering 10 subjects in English and Chinese. Unlike traditional model-based evaluations, MDEval provides excellent interpretability by combining model-based generation tasks and statistical methods. Our results demonstrate that MDEval achieves a Spearman correlation of 0.791 and an accuracy of 84.1% with human, outperforming existing methods by a large margin. Extensive experimental results also show that through fine-tuning over our proposed dataset, less performant open-source models are able to achieve comparable performance to GPT-4o in terms of Markdown Awareness. To ensure reproducibility and transparency, MDEval is open sourced at https://github.com/SWUFE-DB-Group/MDEval-Benchmark.

cs.CL↗

arXiv2025

A Semi-Analytic model for Effects of Fuzzy Dark Matter Granule Perturbations on Orbital Motion

In fuzzy dark matter scenarios, the quantum wave nature of ultralight axion-like particles generates stochastic density fluctuations inside dark matter halos. These fluctuations, known as granules, perturb the orbits of subhalos and other orbiting bodies. While previous studies have simulated these effects using N-body techniques or modeled them statistically using diffusion approximations, we propose an alternative framework based on representing the perturbations as a Fourier series with random coefficients, which can be applied to individual orbits, not just populations. We extend the model to finite-size subhalos, identifying a critical length scale below which subhalos behave as point-mass particles. In contrast, larger subhalos exhibit suppressed perturbations from granules due to their extended mass profiles. Using FDM-Simulator, we validate our finite-size model by isolating granule accelerations and confirming their statistical effects on subhalo dynamics.

astro-ph.CO↗

arXiv2025

Utilizing Large Language Models for Information Extraction from Real Estate Transactions

Real estate sales contracts contain crucial information for property transactions, but manual data extraction can be time-consuming and error-prone. This paper explores the application of large language models, specifically transformer-based architectures, for automated information extraction from real estate contracts. We discuss challenges, techniques, and future directions in leveraging these models to improve efficiency and accuracy in real estate contract analysis. We generated synthetic contracts using the real-world transaction dataset, thereby fine-tuning the large-language model and achieving significant metrics improvements and qualitative improvements in information retrieval and reasoning tasks.

cs.CL↗

arXiv2025

Symmetry-Enriched Topological Phases and Their Gauging: A String-Net Model Realization

We present a systematic framework for constructing exactly-solvable lattice models of symmetry-enriched topological (SET) phases based on an enlarged version of the string-net model. We also gauge the global symmetries of our SET models to obtain string-net models of pure topological phases. Without invoking externally imposed onsite symmetry actions, our approach promotes the string-net model of a pure topological order, specified by an input unitary fusion category $\mathscr{F}$, to an SET model, specified by a multifusion category together with a set of isomorphisms. Two complementary construction strategies are developed in the main text: (i) promotion via outer automorphisms of $\mathscr{F}$ and (ii) promotion via the Frobenius algebras of $\mathscr{F}$. The global symmetries derived via these two strategies are intrinsic to topological phases and are thus termed blood symmetries, as opposed to adopted symmetries, which can be arbitrarily imposed on topological phases. We propose the concept of symmetry-gauging family of topological phases, which are related by gauging their blood symmetries. With our approach, we construct the first explicit lattice realization of a nonabelian-symmetry-enriched topological phase -- the $S_3$ symmetry-enriched $\mathbb{Z}_2 \times \mathbb{Z}_2$ quantum-double phase. The approach further reveals the role of local excitations in SET phases and establishes their symmetry constraints.

cond-mat.str-el↗

arXiv2025

Application of Vision-Language Model to Pedestrians Behavior and Scene Understanding in Autonomous Driving

Vision-language models (VLMs) have become a promising approach to enhancing perception and decision-making in autonomous driving. The gap remains in applying VLMs to understand complex scenarios interacting with pedestrians and efficient vehicle deployment. In this paper, we propose a knowledge distillation method that transfers knowledge from large-scale vision-language foundation models to efficient vision networks, and we apply it to pedestrian behavior prediction and scene understanding tasks, achieving promising results in generating more diverse and comprehensive semantic attributes. We also utilize multiple pre-trained models and ensemble techniques to boost the model's performance. We further examined the effectiveness of the model after knowledge distillation; the results show significant metric improvements in open-vocabulary perception and trajectory prediction tasks, which can potentially enhance the end-to-end performance of autonomous driving.

cs.CV↗

arXiv2025

Technical Report of TeleChat2, TeleChat2.5 and T1

We introduce the latest series of TeleChat models: \textbf{TeleChat2}, \textbf{TeleChat2.5}, and \textbf{T1}, offering a significant upgrade over their predecessor, TeleChat. Despite minimal changes to the model architecture, the new series achieves substantial performance gains through enhanced training strategies in both pre-training and post-training stages. The series begins with \textbf{TeleChat2}, which undergoes pretraining on 10 trillion high-quality and diverse tokens. This is followed by Supervised Fine-Tuning (SFT) and Direct Preference Optimization (DPO) to further enhance its capabilities. \textbf{TeleChat2.5} and \textbf{T1} expand the pipeline by incorporating a continual pretraining phase with domain-specific datasets, combined with reinforcement learning (RL) to improve performance in code generation and mathematical reasoning tasks. The \textbf{T1} variant is designed for complex reasoning, supporting long Chain-of-Thought (CoT) reasoning and demonstrating substantial improvements in mathematics and coding. In contrast, \textbf{TeleChat2.5} prioritizes speed, delivering rapid inference. Both flagship models of \textbf{T1} and \textbf{TeleChat2.5} are dense Transformer-based architectures with 115B parameters, showcasing significant advancements in reasoning and general task performance compared to the original TeleChat. Notably, \textbf{T1-115B} outperform proprietary models such as OpenAI's o1-mini and GPT-4o. We publicly release \textbf{TeleChat2}, \textbf{TeleChat2.5} and \textbf{T1}, including post-trained versions with 35B and 115B parameters, to empower developers and researchers with state-of-the-art language models tailored for diverse applications.

cs.CL↗

arXiv2025

Dark Side of Modalities: Reinforced Multimodal Distillation for Multimodal Knowledge Graph Reasoning

The multimodal knowledge graph reasoning (MKGR) task aims to predict the missing facts in the incomplete MKGs by leveraging auxiliary images and descriptions of entities. Existing approaches are trained with single-target objectives, which neglect the probabilistic correlations of entity labels, especially in non-target entities. Moreover, previous studies incorporate all modalities statically or adaptively, overlooking the negative impacts of irrelevant or misleading information in the incompetent modalities. To address these issues, we introduce a novel Reinforced Multimodal Distillation framework, exploiting the Dark Side of Modalities (DSoM) from two perspectives: (1) Dark knowledge from non-target entities: We propose to train a unimodal KGR model through logit distillation to mimic the multimodal soft labels provided by pre-trained multimodal teacher models. The multimodal soft labels could provide rich supervision signals with subtle correlations among both target and non-target entities from multiple perspectives. We further decouple logits into neighbor entities and non-neighbor entities to divide into two types of correlations. (2) Dark side in unhelpful modalities: To exclude the adverse effects of unhelpful modalities, we introduce a reinforced teacher combination mechanism that dynamically selects the optimal set of multimodal teachers for each triple. The agent is trained to maximize the rewards, which are only assigned to the beneficial multimodal combination strategies for the student model. Comprehensive experiments demonstrate the effectiveness of DSoM framework on 5 MKGR datasets. Codes are available at github.com/OreOZhao/DSoM.

cs.MM↗

arXiv2025

The THESAN project: environmental drivers of Local Group reionization

The timing of cosmic reionization across Local Group (LG) analogues provides insights into their early histories and surrounding large-scale structure. Using the radiation-hydrodynamic simulation THESAN-1 and its dark matter-only counterpart THESAN-DARK-1, we track the reionization histories of all haloes, including 224 LG analogues within the proximity of any of the 20 Virgo-like clusters with halo masses above 10^14 Msun at z=0 and their environments. The statistically controlled samples quantify how the reionization redshift (z_reion) correlates with halo mass, local overdensity, and present-day pair properties. Even at fixed mass, haloes in denser regions ionize earlier, and increasing the overdensity smoothing scale systematically suppresses small-scale structure, including local variations and environmental gradients in z_reion. Virgo-like clusters accelerate reionization in their surroundings out to ~5-10 cMpc, beyond which local overdensity again becomes the dominant factor. Within LG pairs, reionization timing offsets reach up to ~150 Myr and correlate with present-day halo separation, reflecting sensitivity to large-scale structure rather than mass ratio in driving asynchronous reionization. The results support an extreme inside-out picture where clustered sources rapidly ionize their immediate neighborhoods, while lower-density regions self-ionize later and voids wait for external homogenization. These links between environment and reionization timing explain the influence of protoclusters and help interpret fossil records in LG dwarfs around the Milky Way. For Milky Way analogues, we find a reionization redshift as early (late) as z_reion = 12.7^{+2.0}_{-1.7} (8.88^{+0.66}_{-0.70}) when considered on 125 ckpc (500 ckpc) scales, with LG analogues following an inside-out reionization picture.

astro-ph.CO↗

arXiv2025

Marco-Bench-MIF: On Multilingual Instruction-Following Capability of Large Language Models

Instruction-following capability has become a major ability to be evaluated for Large Language Models (LLMs). However, existing datasets, such as IFEval, are either predominantly monolingual and centered on English or simply machine translated to other languages, limiting their applicability in multilingual contexts. In this paper, we present an carefully-curated extension of IFEval to a localized multilingual version named Marco-Bench-MIF, covering 30 languages with varying levels of localization. Our benchmark addresses linguistic constraints (e.g., modifying capitalization requirements for Chinese) and cultural references (e.g., substituting region-specific company names in prompts) via a hybrid pipeline combining translation with verification. Through comprehensive evaluation of 20+ LLMs on our Marco-Bench-MIF, we found that: (1) 25-35% accuracy gap between high/low-resource languages, (2) model scales largely impact performance by 45-60% yet persists script-specific challenges, and (3) machine-translated data underestimates accuracy by7-22% versus localized data. Our analysis identifies challenges in multilingual instruction following, including keyword consistency preservation and compositional constraint adherence across languages. Our Marco-Bench-MIF is available at https://github.com/AIDC-AI/Marco-Bench-MIF.

cs.CL↗