arXiv subjects

Jun Zhang

Publications and source records attributed to Jun Zhang.

At least 235 records · Page 13Linked to original sources

arXiv2026

Puzzle it Out: Local-to-Global World Model for Offline Multi-Agent Reinforcement Learning

Offline multi-agent reinforcement learning (MARL) aims to solve cooperative decision-making problems in multi-agent systems using pre-collected datasets. Existing offline MARL methods primarily constrain training within the dataset distribution, resulting in overly conservative policies that struggle to generalize beyond the support of the data. While model-based approaches offer a promising solution by expanding the original dataset with synthetic data generated from a learned world model, the high dimensionality, non-stationarity, and complexity of multi-agent systems make it challenging to accurately estimate the transitions and reward functions in offline MARL. Given the difficulty of directly modeling joint dynamics, we propose a local-to-global (LOGO) world model, a novel framework that leverages local predictions-which are easier to estimate-to infer global state dynamics, thus improving prediction accuracy while implicitly capturing agent-wise dependencies. Using the trained world model, we generate synthetic data to augment the original dataset, expanding the effective state-action space. To ensure reliable policy learning, we further introduce an uncertainty-aware sampling mechanism that adaptively weights synthetic data by prediction uncertainty, reducing approximation error propagation to policies. In contrast to conventional ensemble-based methods, our approach requires only an additional encoder for uncertainty estimation, significantly reducing computational overhead while maintaining accuracy. Extensive experiments across 8 scenarios against 8 baselines demonstrate that our method surpasses state-of-the-art baselines on standard offline MARL benchmarks, establishing a new model-based baseline for generalizable offline multi-agent learning.

cs.AI↗

arXiv2026

Recurrent Preference Memory for Efficient Long-Sequence Generative Recommendation

Generative recommendation (GenRec) models typically model user behavior via full attention, but scaling to lifelong sequences is hindered by prohibitive computational costs and noise accumulation from stochastic interactions. To address these challenges, we introduce Rec2PM, a framework that compresses long user interaction histories into compact Preference Memory tokens. Unlike traditional recurrent methods that suffer from serial training, Rec2PM employs a novel self-referential teacher-forcing strategy: it leverages a global view of the history to generate reference memories, which serve as supervision targets for parallelized recurrent updates. This allows for fully parallel training while maintaining the capability for iterative updates during inference. Additionally, by representing memory as token embeddings rather than extensive KV caches, Rec2PM achieves extreme storage efficiency. Experiments on large-scale benchmarks show that Rec2PM significantly reduces inference latency and memory footprint while achieving superior accuracy compared to full-sequence models. Analysis reveals that the Preference Memory functions as a denoising Information Bottleneck, effectively filtering interaction noise to capture robust long-term interests.

cs.IR↗

arXiv2026

Reasoning to Rank: An End-to-End Solution for Exploiting Large Language Models for Recommendation

Recommender systems are tasked to infer users' evolving preferences and rank items aligned with their intents, which calls for in-depth reasoning beyond pattern-based scoring. Recent efforts start to leverage large language models (LLMs) for recommendation, but how to effectively optimize the model for improved recommendation utility is still under explored. In this work, we propose Reasoning to Rank, an end-to-end training framework that internalizes recommendation utility optimization into the learning of step-by-step reasoning in LLMs. To avoid position bias in LLM reasoning and enable direct optimization of the reasoning process, our framework performs reasoning at the user-item level and employs reinforcement learning for end-to-end training of the LLM. Experiments on three Amazon datasets and a large-scale industrial dataset showed consistent gains over strong conventional and LLM-based solutions. Extensive in-depth analyses validate the necessity of the key components in the proposed framework and shed lights on the future developments of this line of work.

cs.IR↗

arXiv2026

GPR: Towards a Generative Pre-trained One-Model Paradigm for Large-Scale Advertising Recommendation

As an intelligent infrastructure connecting users with commercial content, advertising recommendation systems play a central role in information flow and value creation within the digital economy. However, existing multi-stage advertising recommendation systems suffer from objective misalignment and error propagation, making it difficult to achieve global optimality, while unified generative recommendation models still struggle to meet the demands of practical industrial applications. To address these issues, we propose GPR (Generative Pre-trained Recommender), the first one-model framework that redefines advertising recommendation as an end-to-end generative task, replacing the traditional cascading paradigm with a unified generative approach. To realize GPR, we introduce three key innovations spanning unified representation, network architecture, and training strategy. First, we design a unified input schema and tokenization method tailored to advertising scenarios, mapping both ads and organic content into a shared multi-level semantic ID space, thereby enhancing semantic alignment and modeling consistency across heterogeneous data. Second, we develop the Heterogeneous Hierarchical Decoder (HHD), a dual-decoder architecture that decouples user intent modeling from ad generation, achieving a balance between training efficiency and inference flexibility while maintaining strong modeling capacity. Finally, we propose a multi-stage joint training strategy that integrates Multi-Token Prediction (MTP), Value-Aware Fine-Tuning and the Hierarchy Enhanced Policy Optimization (HEPO) algorithm, forming a complete generative recommendation pipeline that unifies interest modeling, value alignment, and policy optimization. GPR has been fully deployed in the Tencent Weixin Channels advertising system, delivering significant improvements in key business metrics including GMV and CTCVR.

cs.IR↗

arXiv2026

DiffuReason: Bridging Latent Reasoning and Generative Refinement for Sequential Recommendation

Latent reasoning has emerged as a promising paradigm for sequential recommendation, enabling models to capture complex user intent through multi-step deliberation. Yet existing approaches often rely on deterministic latent chains that accumulate noise and overlook the uncertainty inherent in user intent, and they are typically trained in staged pipelines that hinder joint optimization and exploration. To address these challenges, we propose DiffuReason, a unified "Think-then-Diffuse" framework for sequential recommendation. It integrates multi-step Thinking Tokens for latent reasoning, diffusion-based refinement for denoising intermediate representations, and end-to-end Group Relative Policy Optimization (GRPO) alignment to optimize for ranking performance. In the Think stage, the model generates Thinking Tokens that reason over user history to form an initial intent hypothesis. In the Diffuse stage, rather than treating this hypothesis as the final output, we refine it through a diffusion process that models user intent as a probabilistic distribution, providing iterative denoising against reasoning noise. Finally, GRPO-based reinforcement learning enables the reasoning and refinement modules to co-evolve throughout training, without the constraints of staged optimization. Extensive experiments on four benchmarks demonstrate that DiffuReason consistently improves diverse backbone architectures. Online A/B tests on a large-scale industrial platform further validate its practical effectiveness.

cs.IR↗

arXiv2026

Internalizing Multi-Agent Reasoning for Accurate and Efficient LLM-based Recommendation

Large Language Models (LLMs) are reshaping recommender systems by leveraging extensive world knowledge and semantic reasoning to interpret user intent. However, effectively integrating these capabilities with collaborative signals while avoiding prohibitive inference latency remains a critical bottleneck. To address this, we propose a trajectory-driven internalization framework to develop a Single-agent Trajectory-Aligned Recommender (STAR). Specifically, to internalize complex reasoning capabilities into a single efficient model, we first design a multi-agent teacher system capable of multi-turn tool usage and reflection. This teacher utilizes a Collaborative Signal Translation mechanism to explicitly convert latent behavioral patterns into descriptive natural language evidence to enhance reasoning accuracy. Subsequently, a trajectory-driven distillation pipeline transfers this agentic logic, including planning, tool usage, and self-reflection, into the compact STAR model. Extensive experiments demonstrate that STAR surpasses its teacher by 8.7% to 39.5% while eliminating iterative latency, paving the way for real-time, reasoning-enhanced recommendation.

cs.IR↗

arXiv2026

Spend Search Where It Pays: Value-Guided Structured Sampling and Optimization for Generative Recommendation

Generative recommendation via autoregressive models has unified retrieval and ranking into a single conditional generation framework. However, fine-tuning these models with Reinforcement Learning (RL) often suffers from a fundamental probability-reward mismatch. Conventional likelihood-dominated decoding (e.g., beam search) exhibits a myopic bias toward locally probable prefixes, which causes two critical failures: (1) insufficient exploration, where high-reward items in low-probability branches are prematurely pruned and rarely sampled, and (2) advantage compression, where trajectories sharing high-probability prefixes receive highly correlated rewards with low within-group variance, yielding a weak comparative signal for RL. To address these challenges, we propose V-STAR, a Value-guided Sampling and Tree-structured Advantage Reinforcement framework. V-STAR forms a self-evolving loop via two synergistic components. First, a Value-Guided Efficient Decoding (VED) is developed to identify decisive nodes and selectively deepen high-potential prefixes. This improves exploration efficiency without exhaustive tree search. Second, we propose Sibling-GRPO, which exploits the induced tree topology to compute sibling-relative advantages and concentrates learning signals on decisive branching decisions. Extensive experiments on both offline and online datasets demonstrate that V-STAR outperforms state-of-the-art baselines, delivering superior accuracy and candidate-set diversity under strict latency constraints.

cs.AI↗

arXiv2026

ChainRec: An Agentic Recommender Learning to Route Tool Chains for Diverse and Evolving Interests

Large language models (LLMs) are increasingly integrated into recommender systems, motivating recent interest in agentic and reasoning-based recommendation. However, most existing approaches still rely on fixed workflows, applying the same reasoning procedure across diverse recommendation scenarios. In practice, user contexts vary substantially-for example, in cold-start settings or during interest shifts, so an agent should adaptively decide what evidence to gather next rather than following a scripted process. To address this, we propose ChainRec, an agentic recommender that uses a planner to dynamically select reasoning tools. ChainRec builds a standardized Tool Agent Library from expert trajectories. It then trains a planner using supervised fine-tuning and preference optimization to dynamically select tools, decide their order, and determine when to stop. Experiments on AgentRecBench across Amazon, Yelp, and Goodreads show that ChainRec consistently improves Avg HR@{1,3,5} over strong baselines, with especially notable gains in cold-start and evolving-interest scenarios. Ablation studies further validate the importance of tool standardization and preference-optimized planning.

cs.IR↗

arXiv2026

Device-independent quantum key distribution over 100 km with single atoms

Device-independent quantum key distribution (DI-QKD) is a key application of the quantum internet. We report the realization of DI-QKD between two single-atom nodes linked by 100-km fibers. To improve the entangling rate, single-photon interference is leveraged for entanglement heralding, and quantum frequency conversion is used to reduce fiber loss. A tailored Rydberg-based emission scheme suppresses the photon recoil effect on the atom without introducing noise. We achieved high-fidelity atom-atom entanglement and positive asymptotic key rates for fiber lengths up to 100 km. At 11 km, 1.2 million heralded Bell pairs were prepared over 624 hours, yielding an estimated extractable finite-size secure key rate of 0.112 bits per event against general attacks. Our results close the gap between proof-of-principle quantum network experiments and real-world applications.

quant-ph↗

arXiv2026

Global center of polynomial Newton system and its non-isochronicity

Using a new compactification (toroidal compactification) and desingularization, we obtain a complete characterization of monodromy at infinity for polynomial Newton system of arbitrary degree, in which we establish an equivalence between the monodromy and the non-existence of 1/2-fractional formal invariant curves. Combining the complete characterization with either Darboux integrability or algebraic reducibility of local centers, we obtain conditions for all cases of global center. Furthermore, investigating the asymptotic behavior of the period function of orbits near infinity, we prove the non-isochronicity for the global center, which consequently solves an open problem proposed by Conti.

math.DS↗

arXiv2026

A class of pseudorandom sequences From Function Fields

Motivated by the constructions of pseudorandom sequences over the cyclic elliptic function fields by Hu \textit{et al.} in \text{[IEEE Trans. Inf. Theory, 53(7), 2007]} and the constructions of low-correlation, large linear span binary sequences from function fields by Xing \textit{et al.} in \text{[IEEE Trans. Inf. Theory, 49(6), 2003]}, we utilize the bound derived by Weil \text{[Basic Number Theory, Grund. der Math. Wiss., Bd 144]} and Deligne \text{[ Lecture Notes in Mathematics, vol. 569 (Springer, Berlin, 1977)]} for the exponential sums over the general algebraic function fields and study the periods, linear complexities, linear complexity profiles, distributions of $r-$patterns, period correlation and nonlinear complexities for a class of $p-$ary sequences that generalize the constructions in \text{[IEEE Trans. Inf. Theory, 49(6), 2003]} and [IEEE Trans. Inf. Theory, 53(7), 2007].

cs.IT↗

arXiv2026

DreamDojo: A Generalist Robot World Model from Large-Scale Human Videos

Being able to simulate the outcomes of actions in varied environments will revolutionize the development of generalist agents at scale. However, modeling these world dynamics, especially for dexterous robotics tasks, poses significant challenges due to limited data coverage and scarce action labels. As an endeavor towards this end, we introduce DreamDojo, a foundation world model that learns diverse interactions and dexterous controls from 44k hours of egocentric human videos. Our data mixture represents the largest video dataset to date for world model pretraining, spanning a wide range of daily scenarios with diverse objects and skills. To address the scarcity of action labels, we introduce continuous latent actions as unified proxy actions, enhancing interaction knowledge transfer from unlabeled videos. After post-training on small-scale target robot data, DreamDojo demonstrates a strong understanding of physics and precise action controllability. We also devise a distillation pipeline that accelerates DreamDojo to a real-time speed of 10.81 FPS and further improves context consistency. Our work enables several important applications based on generative world models, including live teleoperation, policy evaluation, and model-based planning. Systematic evaluation on multiple challenging out-of-distribution (OOD) benchmarks verifies the significance of our method for simulating open-world, contact-rich tasks, paving the way for general-purpose robot world models.

cs.RO↗

arXiv2026

Dynamics of string breaking and revival in a Rydberg atomic chain

String breaking is one of the most representative non-perturbative dynamics processes in confinement theory, typically associated with the creation of particle-antiparticle pairs. In this paper, we take a one-dimensional Rydberg atomic chain to theoretically study the dynamical of finite-length string state. Under different string tension conditions, we find that the string dynamics exhibits two clearly distinguishable evolution characteristics: one is that the string breaks and the system enters a superposition state space containing multiple meson state configurations; the other is localized string dynamics, in which the string undergoes local breaking but can then recombine and return to a state close to the initial structure, with the breaking and recombination processes recurring over a long time scale. Through the analysis of the evolution of different meson state configurations, we visually depict the redistribution of configuration weights during the string breaking process, and reveal the observable recovery characteristics of the string after breaking. Further analysis shows that the enhancement of quantum fluctuations can increase the weight of the double-meson state configurations in the system wave function without changing the dominant dynamical behavior. The above results present a rich picture of string breaking dynamics in a one-dimensional Rydberg atomic chain and provide insights for studying confinement physics and related gauge field theory phenomena on quantum simulation platforms.

cond-mat.quant-gas↗

arXiv2026

GAS: Enhancing Reward-Cost Balance of Generative Model-assisted Offline Safe RL

Offline Safe Reinforcement Learning (OSRL) aims to learn a policy to achieve high performance in sequential decision-making while satisfying constraints, using only pre-collected datasets. Recent works, inspired by the strong capabilities of Generative Models (GMs), reformulate decision-making in OSRL as a conditional generative process, where GMs generate desirable actions conditioned on predefined reward and cost values. However, GM-assisted methods face two major challenges in OSRL: (1) lacking the ability to "stitch" optimal transitions from suboptimal trajectories within the dataset, and (2) struggling to balance reward targets with cost targets, particularly when they are conflict. To address these issues, we propose Goal-Assisted Stitching (GAS), a novel algorithm designed to enhance stitching capabilities while effectively balancing reward maximization and constraint satisfaction. To enhance the stitching ability, GAS first augments and relabels the dataset at the transition level, enabling the construction of high-quality trajectories from suboptimal ones. GAS also introduces novel goal functions, which estimate the optimal achievable reward and cost goals from the dataset. These goal functions, trained using expectile regression on the relabeled and augmented dataset, allow GAS to accommodate a broader range of reward-cost return pairs and achieve a better tradeoff between reward maximization and constraint satisfaction compared to human-specified values. The estimated goals then guide policy training, ensuring robust performance under constrained settings. Furthermore, to improve training stability and efficiency, we reshape the dataset to achieve a more uniform reward-cost return distribution. Empirical results validate the effectiveness of GAS, demonstrating superior performance in balancing reward maximization and constraint satisfaction compared to existing methods.

cs.LG↗

arXiv2026

Large Language Model as Meta-Surrogate for Data-Driven Many-Task Optimization: A Proof-of-Principle Study

In many-task optimization scenarios, surrogate models are valuable for mitigating the computational burden of repeated fitness evaluations across tasks. This study proposes a novel meta-surrogate framework to assist many-task optimization, by leveraging the knowledge transfer strengths and emergent capabilities of large language models (LLMs). We formulate a unified framework for many-task fitness prediction, by defining a universal model with metadata to fit a group of problems. Fitness prediction is performed on metadata and decision variables, enabling efficient knowledge sharing across tasks and adaptability to new tasks. The LLM-based meta-surrogate treats fitness prediction as conditional probability estimation, employing a unified token sequence representation for task metadata, inputs, and outputs. This approach facilitates efficient inter-task knowledge sharing through shared token embeddings and captures complex task dependencies via multi-task model training. Experimental results demonstrate the model's emergent generalization ability, including zero-shot performance on problems with unseen dimensions. When integrated into evolutionary transfer optimization (ETO), our framework supports dual-level knowledge transfer -- at both the surrogate and individual levels -- enhancing optimization efficiency and robustness. This work establishes a novel foundation for applying LLMs in surrogate modeling, offering a versatile solution for many-task optimization.

cs.LG↗

arXiv2026

Common Envelope Evolution of Ultralight Boson Clouds

Ultralight bosons can be excited around spinning black holes via black hole superradiance. These boson clouds may play an important role in the orbital evolution of binary black holes. In this work, we investigate the formation and evolution of common envelopes of ultralight boson clouds in comparable mass-ratio black hole binaries. We describe the cloud evolution using gravitational molecular eigenstates and analyze the possible level transitions during orbital decay, as well as the impact on orbital dynamics. We find that the cloud can generally lead to eccentricity growth. In particular, the eccentricity may vary significantly during level transition, leaving an eccentricity of ${\cal O}(0.1)$ within the detection band of ground-based gravitational wave detectors.

gr-qc↗

arXiv2026

Focus-dLLM: Accelerating Long-Context Diffusion LLM Inference via Confidence-Guided Context Focusing

Diffusion Large Language Models (dLLMs) deliver strong long-context processing capability in a non-autoregressive decoding paradigm. However, the considerable computational cost of bidirectional full attention limits the inference efficiency. Although sparse attention is promising, existing methods remain ineffective. This stems from the need to estimate attention importance for tokens yet to be decoded, while the unmasked token positions are unknown during diffusion. In this paper, we present Focus-dLLM, a novel training-free attention sparsification framework tailored for accurate and efficient long-context dLLM inference. Based on the finding that token confidence strongly correlates across adjacent steps, we first design a past confidence-guided indicator to predict unmasked regions. Built upon this, we propose a sink-aware pruning strategy to accurately estimate and remove redundant attention computation, while preserving highly influential attention sinks. To further reduce overhead, this strategy reuses identified sink locations across layers, leveraging the observed cross-layer consistency. Experimental results show that our method offers more than $29\times$ lossless speedup under $32K$ context length. The code is publicly available at: https://github.com/Longxmas/Focus-dLLM

cs.CL↗

arXiv2026

Anisotropic time evolution of sound modes in Bjorken expanding holographic plasma

The speed of sound is a key parameter for characterizing equilibrium states. However, sound waves change their properties when propagating through rapidly evolving anisotropic media, such as the quark-gluon plasma created in heavy-ion collisions. This paper uses $\mathcal{N}=4$ Super-Yang-Mills theory to numerically study the time evolution of the speed and attenuation of sound modes along with the relaxation time in a plasma undergoing Bjorken expansion from various initial states in a quasi-static approximation. The longitudinal Bjorken expansion breaks the isotropy, resulting in two distinct sound speeds that range from just below the conformal value to the speed of light. An anisotropic hydrodynamic description is constructed and its applicability is discussed. Implications for the analysis of heavy ion data are considered.

nucl-th↗