arXiv subjects

Yunkun Xu

Publications and source records attributed to Yunkun Xu.

6 recordsLinked to original sources

arXiv2026

AgentOmnia: Scaling Agentic Models for Full-Scenario Applications

Large language model agents have advanced rapidly, yet progress remains fragmented across domains, capabilities, task difficulty, and interaction settings. We frame this as full-scenario agentic scaling and present AgentOmnia, a framework coordinating task-space definition, data synthesis, post-training, evaluation, and improvement across To-Consumer (ToC), To-Business (ToB), and To-Employee (ToE) applications. An extensible Domain x Capability x Atomic Difficulty taxonomy aligns these stages and enables fine-grained diagnosis with OmniaBench. AgentOmnia combines bidirectional environment-task synthesis with tool-dependency, program-structured, and solver-based pipelines, constructing 5,018 stateful environments with 255,375 tools and 52,361 tasks. Programs, solvers, and verifiers provide correctness signals, while supervised fine-tuning, online agentic reinforcement learning, and a rollback curriculum support post-training. Evaluation failures translate into Product Requirement Documents (PRDs) for targeted self-evolution. Starting from Qwen3-30B-A3B-Thinking-2507, AgentOmnia raises the pass rate on the OmniaBench challenging subset from 9.16% to 37.11% and the macro-average across OmniaBench, $\tau^2$-Bench, DeepPlanning, and VitaBench from 22.86% to 41.69%. Under a unified protocol,it leads the evaluated agentic post-trained baselines on OmniaBench and retains the highest four-benchmark macro-average. It also surpasses Qwen3-235B-A22B-Thinking-2507 on all four benchmarks and exceeds Qwen3.5-35B-A3B on the macro-average. Gains span three application splits, ten capability dimensions, eight atomic-difficulty factors, and 76 of 90 level-1 domains, indicating broad rather than category-specific improvement. A one-round study provides initial evidence for PRD-guided self-evolution, motivating validation at larger scales and in industrial settings.

cs.AI↗

arXiv2026

SearchArt: Training Long-Horizon Search Agent with Scalable Synthetic and Verified Task

Recent advances in large language models (LLMs) have enabled search agents to autonomously tackle complex tasks across extended search and reasoning horizons. However, training effective search agents remains challenging due to the lack of scalable and long-horizon tasks, and the difficulty of evaluating and correcting intermediate reasoning and tool-use behaviors. We introduce SearchArt, a scalable framework for training long-horizon search agents through verification-driven task synthesis and a multi-stage post-training pipeline. SearchArt constructs large-scale datasets for complex search-, research- and user-oriented tasks by synthesizing diverse information-seeking QA pairs and corresponding search trajectories from web documents and automatically generated evidence graphs. To ensure the reliability of the synthesized data, we design a verification pipeline that jointly evaluates QA consistency, trajectory quality, and the relevance of retrieved evidence. The verified trajectories are subsequently used in a multi-stage training process comprising supervised fine-tuning and reinforcement learning-based policy optimization. Search agents trained with SearchArt exhibit adaptive search planning, iterative evidence aggregation, and complex reasoning over extended interaction horizons. Experimental results demonstrate that, with only (Qwen3.5-) 27B parameters, SearchArt scores 74.39 on BrowseComp-ZH, 70.06 on BrowseComp, and 52.55 on Deepresearch-bench, matching or surpassing frontier closed-source agents on both deepsearch and deepresearch benchmarks.

cs.IR↗

arXiv2026

LLM Evolution as an Industry-Scale Ecosystem: A Lifecycle Perspective on Continual Learning

Continual learning capability is critical for Industrial LLMs, as deployed models must be continuously updated to meet evolving requirements and environments, rather than repeatedly retrained from scratch. However, most existing research focuses on improvements on static benchmarks, failing to capture real industrial needs. In this survey, we reformulate Industrial Continual Learning (ICL) for LLMs as a closed-loop update-and-release problem in a versioned ecosystem, where updates propagate hierarchically to industrial, application-specific models and LLM-powered applications, with capability inheritance and transfer across versions and model families. From this ecosystem perspective, we identify three core challenges: repeated adaptation erodes model plasticity, foundation-model upgrades break capability inheritance, and long-term sustainability is constrained by deployment requirements. We then organize the technical landscape of ICL around five lifecycle design principles: preserving plasticity headroom, treating upgrades as capability transfer, enabling trustworthy continual reinforcement learning, making training recipes self-optimizing, and building accountability as a base layer for long-term iteration. For each principle, we synthesize representative technical directions. Finally, we evaluate the maturity of each principle and its technical components via an evidence-based lens, identify key gaps hindering real-world deployment, and outline a practical ICL deployment blueprint and a pathway for feeding industrial realities back into academic research.

cs.LG↗

arXiv2025

Gradient-Adaptive Policy Optimization: Towards Multi-Objective Alignment of Large Language Models

Reinforcement Learning from Human Feedback (RLHF) has emerged as a powerful technique for aligning large language models (LLMs) with human preferences. However, effectively aligning LLMs with diverse human preferences remains a significant challenge, particularly when they are conflict. To address this issue, we frame human value alignment as a multi-objective optimization problem, aiming to maximize a set of potentially conflicting objectives. We introduce Gradient-Adaptive Policy Optimization (GAPO), a novel fine-tuning paradigm that employs multiple-gradient descent to align LLMs with diverse preference distributions. GAPO adaptively rescales the gradients for each objective to determine an update direction that optimally balances the trade-offs between objectives. Additionally, we introduce P-GAPO, which incorporates user preferences across different objectives and achieves Pareto solutions that better align with the user's specific needs. Our theoretical analysis demonstrates that GAPO converges towards a Pareto optimal solution for multiple objectives. Empirical results on Mistral-7B show that GAPO outperforms current state-of-the-art methods, achieving superior performance in both helpfulness and harmlessness.

cs.CL↗

arXiv2024

Strain-Enabled Giant Second-Order Susceptibility in Monolayer WSe$_2$

Monolayer WSe$_2$ (ML WSe$_2$) exhibits a high second-harmonic generation (SHG) efficiency under single 1-photon (1-p) or 2-photon (2-p) resonant excitation conditions due to enhanced second-order susceptibility compared with off-resonance excitation states \cite{lin2021narrow,wang2015giant}. Here, we propose a novel strain engineering approach to dramatically boost the in-plane second-order nonlinear susceptibility ($\chi_{yyy}$ ) of ML WSe$_2$ by tuning the biaxial strain to shift two K-valley excitons (the A-exciton and a high-lying exciton (HX)) into double resonance. We first identify the A-exciton and HX from the 2D Mott-Wannier model for pristine ML WSe$_2$ and calculate the $\chi_{yyy}$ under either 1-p or 2-p resonance excitations, and observe a $\sim$ 39-fold $\chi_{yyy}$ enhancement arising from the 2-p HX resonance state compared with the A-exciton case. By applying a small uniform biaxial strain (0.16\%), we observe an exciton double resonance state ($E_{\rm{HX}}$ = 2$E_{\rm{A}}$, $E_{\rm{HX}}$ and $E_{\rm{A}}$ are the exciton absorption energies), which yields up to an additional 52-fold enhancement in $\chi_{yyy}$ compared to the 2-p HX resonance state, indicating an overall $\sim$ 2000-fold enhancement compared to the single 2-p A-exciton resonance state reported in Ref \cite{wang2015giant}. Further exploration of the strain-engineered exciton states (with biaxial strain around 0.16\%) reveals that double resonance also occurs at other wavevectors near the K valley, leading to other enhancement states in $\chi_{yyy}$, confirming that strain engineering is an effective approach for enhancing $\chi_{yyy}$. Our findings suggest new avenues for strain engineering the optical properties of 2D materials for novel nonlinear optoelectronic applications.

physics.optics↗

arXiv2021

Look Before You Leap: Safe Model-Based Reinforcement Learning with Human Intervention

Safety has become one of the main challenges of applying deep reinforcement learning to real world systems. Currently, the incorporation of external knowledge such as human oversight is the only means to prevent the agent from visiting the catastrophic state. In this paper, we propose MBHI, a novel framework for safe model-based reinforcement learning, which ensures safety in the state-level and can effectively avoid both "local" and "non-local" catastrophes. An ensemble of supervised learners are trained in MBHI to imitate human blocking decisions. Similar to human decision-making process, MBHI will roll out an imagined trajectory in the dynamics model before executing actions to the environment, and estimate its safety. When the imagination encounters a catastrophe, MBHI will block the current action and use an efficient MPC method to output a safety policy. We evaluate our method on several safety tasks, and the results show that MBHI achieved better performance in terms of sample efficiency and number of catastrophes compared to the baselines.

cs.AI↗