arXiv subjects

Ai Han

Publications and source records attributed to Ai Han.

7 recordsLinked to original sources

arXiv2026

ToolAnchor: Anchoring Counterfactual Context to Boost Agentic Tool-use Capability

Tool-augmented large language model agents excel at long-horizon tasks, yet they are typically post-trained on fixed toolsets. When tasks demand new tools, these agents struggle to incorporate them effectively, and retraining from scratch is often impractical. We identify the core obstacle in such toolset expansion problem as behavioral inertia: the tendency of agents to fall back on familiar tools and established reasoning patterns despite having access to new ones. We demonstrate that injecting counterfactual anchor contexts at critical decision points can break this inertia, recovering failed trajectories by eliciting suppressed agent capabilities. To scale this insight, we propose ToolAnchor, a framework that uses teacher models to hypothesize these counterfactual contexts, verifies them via student rollouts, and internalizes the successful interventions through agentic post-training. Extensive evaluations across general AI assistant (GAIA), textual search (BrowseComp), and visual search (VDR-Bench) tasks demonstrate that ToolAnchor consistently exhibits competitive performance under expanded toolsets. Our work bridges the gap between static post-training and dynamic adaptation, charting a new path for scalable agentic reinforcement learning.

cs.AI↗

arXiv2026

Hierarchical Attacks for Multi-Modal Multi-Agent Reasoning

Multi-modal multi-agent systems (MM-MAS) have gained increasing attention for their capacity to enable complex reasoning and coordination across diverse modalities. As these systems continue to expand in scale and functionality, investigating their potential vulnerabilities has become increasingly important. However, existing studies on adversarial attacks in multi-agent systems primarily focus on isolated agents or unimodal settings, leaving the vulnerabilities of MM-MAS largely underexplored. To bridge this gap, we introduce HAM$^{3}$, a Hierarchical Attack framework for multi-modal multi-agent systems that decomposes attacks into three interconnected layers. Specifically, at the perception layer, HAM$^{3}$ mounts attacks by perturbing visual inputs, textual inputs, and their fused visual-textual representations. At the communication layer, it performs communication-level attacks that corrupt message content and interaction topology, such as manipulating shared context or communication links to distort collective information flow. At the reasoning layer, it conducts reasoning-level attacks that interfere with each agent's cognitive pipeline, biasing reasoning trajectories and ultimately compromising final decisions. We evaluate HAM$^{3}$ on the GQA benchmark through multi-agent systems built on distinct reasoning paradigms including ReAct, Plan-and-Solve, and Reflexion. Experiments demonstrate that our framework achieves an Attack Success Rate of up to 78.3%, with reasoning-layer attacks being the most effective. More than half of the successful attacks lead multiple agents to produce consistent errors. These findings offer valuable insights for building more robust and interpretable multi-agent intelligence.

cs.AI↗

arXiv2026

OxyGent: Making Multi-Agent Systems Modular, Observable, and Evolvable via Oxy Abstraction

Deploying production-ready multi-agent systems (MAS) in complex industrial environments remains challenging due to limitations in scalability, observability, and autonomous evolution. We present OxyGent, an open-source framework driven by two core novelties: a unified Oxy abstraction and the OxyBank evolution engine. The unified abstraction encapsulates agents, tools, LLMs, and reasoning flows as pluggable atomic components, enabling Lego-like scalable system composition and non-intrusive monitoring. To enhance observability, OxyGent introduces permission-driven dynamic planning that replaces rigid workflows with execution graphs generated at runtime, providing adaptive visualizations. Furthermore, to support continuous evolution, OxyBank serves as an AI asset management platform that drives automated data backflow, annotation, and joint evolution. Empirical evaluations and real-world case studies show that OxyGent provides a robust and scalable foundation for MAS. OxyGent is fully open-sourced under the Apache License 2.0 at https://github.com/jd-opensource/OxyGent.

cs.AI↗

arXiv2026

Rollout-Training Co-Design for Efficient LLM-Based Multi-Agent Reinforcement Learning

Despite algorithm-level innovations for multi-agent reinforcement learning (MARL), the underlying networked infrastructure for large-scale MARL training remains underexplored. Existing training frameworks primarily optimize for single-agent scenarios and fail to address the unique system-level challenges of MARL, including rollout-training synchronization barriers, rollout load imbalance, and training resource underutilization. To bridge this gap, we propose FlexMARL, the first end-to-end training framework that holistically optimizes rollout, training, and their orchestration for large-scale LLM-based MARL. Specifically, FlexMARL introduces the joint orchestrator to manage data flow under the rollout-training disaggregated architecture. Building upon the experience store, a novel micro-batch driven asynchronous pipeline eliminates the synchronization barriers while providing strong consistency guarantees. Rollout engine adopts a parallel sampling scheme combined with hierarchical load balancing, which adapts to skewed inter/intra-agent request patterns. Training engine achieves on-demand hardware binding through agent-centric resource allocation. The training states of different agents are swapped via unified and location-agnostic communication. Empirical results on a large-scale production cluster demonstrate that FlexMARL achieves up to 7.3x speedup and improves hardware utilization by up to 5.6x compared to existing frameworks.

cs.LG↗

arXiv2025

NaviAgent: Graph-Driven Bilevel Planning for Scalable Tool Orchestration

Large Language Models (LLMs) increasingly act as function-call agents that invoke external tools to tackle tasks beyond their static knowledge. However, they typically invoke tools one at a time without a global view of task structure. As tools often depend on one another, this leads to error accumulation and poor scalability, particularly when scaling to hundreds or thousands of tools. To address these limitations, we propose NaviAgent, an explicit bilevel architecture that decouples task planning from tool execution through graph-based modeling of tool relations. At the planning level, the LLM-based agent decides whether to respond directly, clarify intent, or retrieve and execute a toolchain independent of inter-tool complexity. At the execution level, a Tool World Navigation Model (TWNM) encodes structural and behavioral relations among tools, steering the agent to compose scalable and robust invocation sequences. Incorporating feedback from real tool interactions, NaviAgent achieves closed-loop alignment between planning and execution, enabling adaptive navigation in large-scale tool ecosystems. Evaluations on API-Bank and ToolBench show consistent improvements in task success rate (TSR), with TWNM yielding an average gain of 13.1 points on complex tasks. Further tests on 50 real APIs across 7 domains show consistent gains of 4.3--12.0 points, with fewer steps and latency, demonstrating robust generalization under real-world dynamics.

cs.AI↗

arXiv2025

HiMA-Ecom: Enabling Joint Training of Hierarchical Multi-Agent E-commerce Assistants

Hierarchical multi-agent systems based on large language models (LLMs) have become a common paradigm for building AI assistants in vertical domains such as e-commerce, where a master agent coordinates multiple specialized sub-agents. Despite their practical importance, realistic benchmarks for training and evaluating such systems remain scarce, and joint optimization across functionally distinct agents is still challenging. To address this gap, we introduce HiMA-Ecom, the first hierarchical multi-agent benchmark tailored for e-commerce scenarios. HiMA-Ecom contains 22.8K instances, including agent-specific supervised fine-tuning samples with memory and system-level input-output pairs for joint multi-agent reinforcement learning. Building upon it, a joint training method named HiMA-R1 is proposed. It presents Variance-Reduction Group Relative Policy Optimization (VR-GRPO), which employs initial trajectory-based Monte Carlo sampling to mitigate the exponential joint action space and selects informative agent groups for efficient updates based on reward variance. Furthermore, an adaptive memory evolution mechanism that repurposes GRPO rewards as cost-free supervisory signals is designed to eliminate repetitive reasoning and accelerate convergence. Experiments on HiMA-Ecom demonstrate that our method, built upon smaller 3B/7B open-source models, achieves performance comparable to that of larger LLMs, such as DeepSeek-R1, and surpasses DeepSeek-V3 by an average of 6\%.

cs.AI↗

arXiv2025

OTPTO: Joint Product Selection and Inventory Optimization in Fresh E-commerce Front-End Warehouses

In China's competitive fresh e-commerce market, optimizing operational strategies, especially inventory management in front-end warehouses, is key to enhance customer satisfaction and to gain a competitive edge. Front-end warehouses are placed in residential areas to ensure the timely delivery of fresh goods and are usually in small size. This brings the challenge of deciding which goods to stock and in what quantities, taking into account capacity constraints. To address this issue, traditional predict-then-optimize (PTO) methods that predict sales and then decide on inventory often don't align prediction with inventory goals, as well as fail to prioritize consumer satisfaction. This paper proposes a multi-task Optimize-then-Predict-then-Optimize (OTPTO) approach that jointly optimizes product selection and inventory management, aiming to increase consumer satisfaction by maximizing the full order fulfillment rate. Our method employs a 0-1 mixed integer programming model OM1 to determine historically optimal inventory levels, and then uses a product selection model PM1 and the stocking model PM2 for prediction. The combined results are further refined through a post-processing algorithm OM2. Experimental results from JD.com's 7Fresh platform demonstrate the robustness and significant advantages of our OTPTO method. Compared to the PTO approach, our OTPTO method substantially enhances the full order fulfillment rate by 4.34% (a relative increase of 7.05%) and narrows the gap to the optimal full order fulfillment rate by 5.27%. These findings substantiate the efficacy of the OTPTO method in managing inventory at front-end warehouses of fresh e-commerce platforms and provide valuable insights for future research in this domain.

cs.LG↗