arXiv subjects

Pengbo Xu

Publications and source records attributed to Pengbo Xu.

At least 19 recordsLinked to original sources

arXiv2026

Synergistic Information Disentanglement for Omni-modal Slide Representation Learning in Computational Pathology

In computational pathology (CPath), developing omni-modal self-supervised learning (SSL) models that integrate histology, genomics, and clinical reports enables transferable representation learning for whole slide images (WSIs). Existing approaches implicitly force heterogeneous modalities into a uniform latent space by contrastive alignment, causing modality collapse where unique, synergistic diagnostic signals (termed as $\mathrm{\Phi}$) are discarded in favor of trivial redundancy. We hypothesize that the strongest task-agnostic SSL training signal stems from distilling the synergistic interactions over merely aligning shared redundancy. To this end, we introduce \textsc{$\mathrm{\Phi}$-Omni}, a synergistic information disentanglement framework grounded in Partial Information Decomposition (PID) theory for slide representation learning. Unlike standard contrastive approaches, \textsc{$\mathrm{\Phi}$-Omni} employs a Synergistic Information Bottleneck (SIB) regulated by the proposed $\mathrm{\Phi}\text{ID}$ objective, which explicitly suppresses marginal redundancy while maximizing irreducible synergy, thereby distilling high-order cross-modal interactions. Following pretraining on breast ($n$=1031) and lung ($n$=919) cohorts, \textsc{$\mathrm{\Phi}$-Omni} demonstrates superior few-shot performance across five independent external datasets spanning eight tasks compared to supervised and SSL baselines. Source code is available here.

cs.CV↗

arXiv2026

From Trajectories to Evidence: Auditable Experimental Records for Industrial Research Agents

Research agents increasingly conduct multi-round machine-learning experiments in industrial recommendation settings and retain the resulting trajectories to guide later decisions. Yet a completed trajectory is not automatically evidence: generated artifacts may be unsupported or incomplete, executed rounds may be invalid or confounded, and later modifications may obscure earlier findings. We study \textbf{trajectory-to-evidence conversion}, asking what a completed research process has actually established. We introduce an evidence-grounded framework that couples bounded verification of consequential artifacts with post-execution claim qualification. A context-isolated generate--verify--repair process checks artifacts for evidence violations and missing downstream requirements before release. After execution, validity and attribution checks consolidate evidence across rounds, qualify intervention-level claims as actionable repairs, diagnostic guards, or withheld findings, and preserve admitted claims as auditable records with explicit provenance and applicability boundaries. A hybrid LLM-assisted controller subsequently applies, defers, or rejects records based on available target evidence. Record audits characterize which claims survive qualification, while downstream diagnostics identify affirmative applicability judgment as a bottleneck for the tested controller. Across paper-to-target adaptations, later rounds often improve on the first, while final rounds frequently underperform an earlier best, exposing non-monotonic trajectory evolution. Candidates produced through the complete workflow also yielded positive online lifts relative to deployed baselines.

cs.IR↗

arXiv2026

Reward Guided Decoding for Generative Recommendation

Generative recommendation formulates recommendation task into an SID sequence autoregressive generation paradigm, but the decoding process is often dominated by generation likelihood. This may conflict with real-world business objectives, where high-value candidates can receive low generation probability and be pruned early during beam search. Existing reranking or training-time alignment methods either intervene too late or require costly model retraining when business preferences change. To this end, we propose \textbf{R}eward \textbf{G}uided \textbf{D}ecoding, named \textbf{RGD}, a controllable decoding framework for industrial value-oriented generative recommendation. We formulate value-guided decoding as a KL-regularized reward maximization problem, deriving a closed-form reward guided decoding distribution that principledly combines generation probability with reward signals. RGD treats the base generator as a reference policy and introduces a reward model as a test-time controller, injecting reward into each decoding step to reshape the search trajectory without retraining the generator. Extensive offline and online experiments demonstrate the effectiveness of our approach for aligning personalization and business value. RGD has been deployed on the Kuaishou platform, bringing consistent improvements in real-world recommendation scenarios.

cs.IR↗

arXiv2026

Unifying Generative Recall and Multi-Objective Ranking in a Single Decoder-Only Sequence

Modern industrial recommendation systems typically separate recall and ranking into two independent stages. Although this cascade supports corpus-level retrieval and fine-grained multi-objective scoring, it causes objective inconsistency, information loss at the candidate hand-off, and redundant user-side context computation. Meanwhile, the generative recall and ranking scaling share a common Transformer-based modeling philosophy, where architectural consistency creates a natural opportunity for unified integration. However, direct sharing remains challenging since the two tasks require different information visibility and optimization methods. Therefore, we propose \textbf{UniR$^2$}, a \textbf{Uni}fied decoder-only Transformer that unifies Generative \textbf{R}ecall and Multi-Objective \textbf{R}anking within a single heterogeneous sequence comprising user context, SID trajectory, and item features. Within this sequence, the generated trajectory serves as a representation bridge between recall and ranking, where Dual-Query Prefix-Causal Attention provides task-specific visibility. The two tasks share the base attention weights but retain separate optimization boundaries, with ranking-side LoRA preserving ranking adaptability without disrupting the generative backbone. Extensive offline experiments on large-scale industrial data demonstrate the effectiveness and efficiency of UniR$^2$ for both recall and ranking. Long-term online A/B tests on Kuaishou platform further show consistent positive gains, validating the practicality of unified model in large-scale recommendation systems.

cs.IR↗

arXiv2026

AgentX: Towards Agent-Driven Self-Iteration of Industrial Recommender Systems

Recommendation algorithm iteration is moving from an artisanal, engineer-bound process toward an industrialized research loop, but this transition remains blocked by a structural execution bottleneck: the idea-to-launch cycle still depends on human engineers to generate hypotheses, modify production code, launch A/B experiments, and attribute online results. Innovation therefore scales linearly with headcount rather than compounding with evidence, compute, and accumulated experimental knowledge. We present AgentX, a production-deployed multi-agent system that fundamentally restructures this production function. AgentX operates as a self-evolving development engine: it autonomously generates, implements, evaluates, and learns from recommendation experiments at a scale and pace that no manual workflow can sustain. The system orchestrates four tightly coupled stages in a closed loop. A Brainstorm Agent synthesizes evidence from historical experiments, system architecture, data analysis, and external research into ranked, executable proposals. A Developing Agent translates each proposal into production-ready code through repository-grounded generation and multi-dimensional reliability verification. An Evaluation Agent conducts safe online rollout with guardrail-vetoed A/B judgment, converting both successes and failures into structured knowledge assets. A Harness Evolution layer (SGPO) then distills execution trajectories into semantic-gradient updates that continuously sharpen the agents themselves -- making the system not merely automated, but self-improving.

cs.AI↗

arXiv2026

MaRI: Accelerating Ranking Model Inference via Structural Re-parameterization in Large Scale Recommendation System

Ranking models, i.e., coarse-ranking and fine-ranking models, serve as core components in large-scale recommendation systems, responsible for scoring massive item candidates based on user preferences. To meet the stringent latency requirements of online serving, structural lightweighting or knowledge distillation techniques are commonly employed for ranking model acceleration. However, these approaches typically lead to a non-negligible drop in accuracy. Notably, the angle of lossless acceleration by optimizing feature fusion matrix multiplication, particularly through structural reparameterization, remains underexplored. In this paper, we propose MaRI, a novel Matrix Re-parameterized Inference framework, which serves as a complementary approach to existing techniques while accelerating ranking model inference without any accuracy loss. MaRI is motivated by the observation that user-side computation is redundant in feature fusion matrix multiplication, and we therefore adopt the philosophy of structural reparameterization to alleviate such redundancy.

cs.IR↗

arXiv2026

OneLive: Dynamically Unified Generative Framework for Live-Streaming Recommendation

Live-streaming recommender system serves as critical infrastructure that bridges the patterns of real-time interactions between users and authors. Similar to traditional industrial recommender systems, live-streaming recommendation also relies on cascade architectures to support large-scale concurrency. Recent advances in generative recommendation unify the multi-stage recommendation process with Transformer-based architectures, offering improved scalability and higher computational efficiency. However, the inherent complexity of live-streaming prevents the direct transfer of these methods to live-streaming scenario, where continuously evolving content, limited lifecycles, strict real-time constraints, and heterogeneous multi-objectives introduce unique challenges that invalidate static tokenization and conventional model framework. To address these issues, we propose OneLive, a dynamically unified generative recommendation framework tailored for live-streaming scenario. OneLive integrates four key components: (i) A Dynamic Tokenizer that continuously encodes evolving real-time live content fused with behavior signal through residual quantization; (ii) A Time-Aware Gated Attention mechanism that explicitly models temporal dynamics for timely decision making; (iii) An efficient decoder-only generative architecture enhanced with Sequential MTP and QK Norm for stable training and accelerated inference; (iv) A Unified Multi-Objective Alignment Framework reinforces policy optimization for personalized preferences.

cs.IR↗

arXiv2025

MurreNet: Modeling Holistic Multimodal Interactions Between Histopathology and Genomic Profiles for Survival Prediction

Cancer survival prediction requires integrating pathological Whole Slide Images (WSIs) and genomic profiles, a challenging task due to the inherent heterogeneity and the complexity of modeling both inter- and intra-modality interactions. Current methods often employ straightforward fusion strategies for multimodal feature integration, failing to comprehensively capture modality-specific and modality-common interactions, resulting in a limited understanding of multimodal correlations and suboptimal predictive performance. To mitigate these limitations, this paper presents a Multimodal Representation Decoupling Network (MurreNet) to advance cancer survival analysis. Specifically, we first propose a Multimodal Representation Decomposition (MRD) module to explicitly decompose paired input data into modality-specific and modality-shared representations, thereby reducing redundancy between modalities. Furthermore, the disentangled representations are further refined then updated through a novel training regularization strategy that imposes constraints on distributional similarity, difference, and representativeness of modality features. Finally, the augmented multimodal features are integrated into a joint representation via proposed Deep Holistic Orthogonal Fusion (DHOF) strategy. Extensive experiments conducted on six TCGA cancer cohorts demonstrate that our MurreNet achieves state-of-the-art (SOTA) performance in survival prediction.

eess.IV↗

arXiv2025

Pantheon: Personalized Multi-objective Ensemble Sort via Iterative Pareto Policy Optimization

In this paper, we provide our milestone ensemble sort work and the first-hand practical experience, Pantheon, which transforms ensemble sorting from a "human-curated art" to a "machine-optimized science". Compared with formulation-based ensemble sort, our Pantheon has the following advantages: (1) Personalized Joint Training: our Pantheon is jointly trained with the real-time ranking model, which could capture ever-changing user personalized interests accurately. (2) Representation inheritance: instead of the highly compressed Pxtrs, our Pantheon utilizes the fine-grained hidden-states as model input, which could benefit from the Ranking model to enhance our model complexity. Meanwhile, to reach a balanced multi-objective ensemble sort, we further devise an \textbf{iterative Pareto policy optimization} (IPPO) strategy to consider the multiple objectives at the same time. To our knowledge, this paper is the first work to replace the entire formulation-based ensemble sort in industry RecSys, which was fully deployed at Kuaishou live-streaming services, serving 400 Million users daily.

cs.SI↗

arXiv2024

Botfip-LLM: An Enhanced Multimodal Scientific Computing Framework Leveraging Knowledge Distillation from Large Language Models

In recent years, the introduction of AI technologies has brought transformative changes to scientific computing. However, AI models typically focus on single-task and single-modal data processing, limiting their application. To address this, multimodal scientific computing frameworks have become a trend. The Botfip framework aligns function images with symbolic operation trees through multimodal training, extracting deep scientific information. However, Botfip struggles with processing Formula Strings, leading to inadequate understanding in multimodal learning. To enhance Botfip's learning of Formula Strings and expand its applicability to related tasks, we propose the Botfip-LLM framework based on knowledge distillation, incorporating pre-trained large language models for aligning symbolic tree data. Experimental analysis shows that the choice of LLM is crucial, with ChatGLM-2 outperforming others in training and testing. Botfip-LLM not only improves performance, generalization, and extrapolation over the original Botfip model but also significantly enhances applicability to Formula String-related tasks, enabling more diverse task handling.

cs.SC↗

arXiv2024

Unleashing the Infinity Power of Geometry: A Novel Geometry-Aware Transformer (GOAT) for Whole Slide Histopathology Image Analysis

The histopathology analysis is of great significance for the diagnosis and prognosis of cancers, however, it has great challenges due to the enormous heterogeneity of gigapixel whole slide images (WSIs) and the intricate representation of pathological features. However, recent methods have not adequately exploited geometrical representation in WSIs which is significant in disease diagnosis. Therefore, we proposed a novel weakly-supervised framework, Geometry-Aware Transformer (GOAT), in which we urge the model to pay attention to the geometric characteristics within the tumor microenvironment which often serve as potent indicators. In addition, a context-aware attention mechanism is designed to extract and enhance the morphological features within WSIs.

eess.IV↗

arXiv2024

YingLong-weather: AI-Based Limited Area Models for Forecasting of Non-precipitation Surface Meteorological Variables

Recently, artificial intelligence-based (AI-based) models for forecasting of global weather have been rapidly developed. Most of the global models are trained on reanalysis datasets with a spatial resolution of 0.25{\deg}*0.25{\deg}. However, research on AI-based high spatial resolution limited area weather forecasting models remains limited. In this study, YingLong, an AI-based limited area weather forecasting model with a spatial resolution of 3 km * 3 km is developed. YingLong employs a parallel structure of global and local blocks to capture multiscale meteorological features and operates much faster than the dynamical limited area model WRF-ARW. In two selected limited areas (one relatively flat and the other featuring significant mountain ranges), YingLong (with lateral boundary condition imposed by the global AI-based model Pangu-weather) demonstrates superior skill in forecasting surface wind speed compared to WRF-ARW. Additionally, it shows comparable skill in forecasting surface temperature and pressure. The accuracy of surface temperature and humidity forecasts can be further improved by applying better boundary conditions. YingLong also addresses issues related to the lateral boundary conditions of AI-based limited area models, such as selecting the width of the lateral boundary region and combining finer and coarser resolution predictions in this region. Therefore, YingLong has a great potential to generate cost-effective multiyear high-resolution synthetic wind speed that maintain meteorological realism both spatially and temporally, aiding in the planning and operations for wind power generation companies.

physics.ao-ph↗

arXiv2024

Bootstrapping OTS-Funcimg Pre-training Model (Botfip) -- A Comprehensive Symbolic Regression Framework

In the field of scientific computing, many problem-solving approaches tend to focus only on the process and final outcome, even in AI for science, there is a lack of deep multimodal information mining behind the data, missing a multimodal framework akin to that in the image-text domain. In this paper, we take Symbolic Regression(SR) as our focal point and, drawing inspiration from the BLIP model in the image-text domain, propose a scientific computing multimodal framework based on Function Images (Funcimg) and Operation Tree Sequence (OTS), named Bootstrapping OTS-Funcimg Pre-training Model (Botfip). In SR experiments, we validate the advantages of Botfip in low-complexity SR problems, showcasing its potential. As a MED framework, Botfip holds promise for future applications in a broader range of scientific computing problems.

cs.SC↗

arXiv2021

L\'{e}vy walk dynamics in non-static media

Almost all the media the particles move in are non-static. Depending on the expected resolution of the studied dynamics and the amplitude of the displacement of the media, sometimes the non-static behaviours of the media can not be ignored. In this paper, we build the model describing L\'evy walks in non-static media, where the physical and comoving coordinates are connected by scale factor. We derive the equation governing the probability density function of the position of the particles in comoving coordinate. Using the Hermite orthogonal polynomial expansions, some statistical properties are obtained, such as mean squared displacements (MSDs) in both coordinates and kurtosis. For some representative non-static media and L\'{e}vy walks, the asymptotic behaviors of MSDs in both coordinates are analyzed in detail. The stationary distributions and mean first passage time for some cases are also discussed through numerical simulations.

cond-mat.stat-mech↗

arXiv2021

Gaussian Process and Levy Walk under Stochastic Non-instantaneous Resetting and Stochastic Rest

A stochastic process with movement, return, and rest phases is considered in this paper. For the movement phase, the particles move following the dynamics of Gaussian process or ballistic type of L\'evy walk, and the time of each movement is random. For the return phase, the particles will move back to the origin with a constant velocity or acceleration or under the action of a harmonic force after each movement, so that this phase can also be treated as a non-instantaneous resetting. After each return, a rest with a random time at the origin follows. The asymptotic behaviors of the mean squared displacements with different kinds of movement dynamics, random resting time, and returning are discussed. The stationary distributions are also considered when the process is localized. Besides, the mean first passage time is considered when the dynamic of movement phase is Brownian motion.

cond-mat.stat-mech↗

arXiv2020

L\'{e}vy walk dynamics in mixed potentials from the perspective of random walk theory

L\'evy walk process is one of the most effective models to describe superdiffusion, which underlies some important movement patterns and has been widely observed in the micro and macro dynamics. From the perspective of random walk theory, here we investigate the dynamics of L\'evy walks under the influences of the constant force field and the one combined with harmonic potential. Utilizing Hermite polynomial approximation to deal with the spatiotemporally coupled analysis challenges, some striking features are detected, including non Gaussian stationary distribution, faster diffusion, and still strongly anomalous diffusion, etc.

cond-mat.stat-mech↗

arXiv2020

L\'evy walk dynamics in an external harmonic potential

L\'evy walks (LWs) are spatiotemporally coupled random-walk processes describing superdiffusive heat conduction in solids, propagation of light in disordered optical materials, motion of molecular motors in living cells, or motion of animals, humans, robots, and viruses. We here investigate a key feature of LWs, their response to an external harmonic potential. In this generic setting for confined motion we demonstrate that LWs equilibrate exponentially and may assume a bimodal stationary distribution. We also show that the stationary distribution has a horizontal slope next to a reflecting boundary placed at the origin, in contrast to correlated superdiffusive processes. Our results generalize LWs to confining forces and settle some long-standing puzzles around LWs.

cond-mat.stat-mech↗

arXiv2019

Continuous time random walks and L\'{e}vy walks with stochastic resetting

Intermittent stochastic processes appear in a wide field, such as chemistry, biology, ecology, and computer science. This paper builds up the theory of intermittent continuous time random walk (CTRW) and L\'{e}vy walk, in which the particles are stochastically reset to a given position with a resetting rate $r$. The mean squared displacements of the CTRW and L\'{e}vy walks with stochastic resetting are calculated, uncovering that the stochastic resetting always makes the CTRW process localized and L\'{e}vy walk diffuse slower. The asymptotic behaviors of the probability density function of L\'evy walk with stochastic resetting are carefully analyzed under different scales of $x$, and a striking influence of stochastic resetting is observed.

cond-mat.stat-mech↗