arXiv subjects

Bo Yang

Publications and source records attributed to Bo Yang.

At least 127 records · Page 7Linked to original sources

arXiv2026

Emerging from Ground: Addressing Intent Deviation in Tool-Using Agents via Deriving Real Calls into Virtual Trajectories

LLMs have advanced tool-using agents for real-world applications, yet they often lead to unexpected behaviors or results. Beyond obvious failures, the subtle issue of "intent deviation" severely hinders reliable evaluation and performance improvement. Existing post-training methods generally leverage either real system samples or virtual data simulated by LLMs. However, the former is costly due to reliance on hand-crafted user requests, while the latter suffers from distribution shift from the real tools in the wild. Additionally, both methods lack negative samples tailored to intent deviation scenarios, hindering effective guidance on preference learning. We introduce RISE, a "Real-to-Virtual" method designed to mitigate intent deviation. Anchoring on verified tool primitives, RISE synthesizes virtual trajectories and generates diverse negative samples through mutation on critical parameters. With synthetic data, RISE fine-tunes backbone LLMs via the two-stage training for intent alignment. Evaluation results demonstrate that data synthesized by RISE achieve promising results in eight metrics covering user requires, execution trajectories and agent responses. Integrating with training, RISE achieves an average 35.28% improvement in Acctask (task completion) and 23.27% in Accintent (intent alignment), outperforming SOTA baselines by 1.20--42.09% and 1.17--54.93% respectively.

cs.AI↗

arXiv2026

ZPD Detector: Data Selection via Capability-Difficulty Alignment for Large Language Models

As the cost of training large language models continues to increase and high-quality training data become increasingly scarce, selecting high-value samples or synthesizing effective training data under limited data budgets has emerged as a critical research problem. Most existing data selection methods rely on static criteria, such as difficulty, uncertainty, or heuristics, and fail to model the evolving relationship between the model and the data. Inspired by the educational theory of the Zone of Proximal Development (ZPD), we propose ZPD Detector, a data selection framework that adopts a bidirectional perspective between models and data by explicitly modeling the alignment between sample difficulty and the model's current capability. ZPD Detector integrates difficulty calibration, model capability estimation based on Item Response Theory (IRT), and a capability-difficulty matching score to dynamically identify the most informative samples at each learning stage, improving data utilization efficiency; moreover, this dynamic matching strategy provides new insights into training strategy design. All code and data will be released after our work be accepted to support reproducible researc

cs.CL↗

arXiv2026

Frontiers of Generative AI for Network Optimization: Theories, Limits, and Visions

While interest in the application of generative AI (GenAI) in network optimization has surged in recent years, its rapid progress has often overshadowed critical limitations intrinsic to generative models that remain insufficiently examined in existing literature. This survey provides a comprehensive review and critical analysis of GenAI in network optimization. We focus on the two dominant paradigms of GenAI including generative diffusion models (GDMs) and large pre-trained models (LPTMs), and organize our discussion around a categorization we introduce, dividing network optimization problems into two primary formulations: one-shot optimization and Markov decision process (MDP). We first trace key works, including foundational contributions from the AI community, and categorize current efforts in network optimization. We also review frontier applications of GDMs and LPTMs in other networking tasks, providing additional context. Furthermore, we present theoretical generalization bounds for GDMs in both one-shot and MDP settings, offering insights into the fundamental factors affecting model performance. Most importantly, we reflect on the overestimated perception of GenAI's general capabilities and caution against the all-in-one illusion it may convey. We highlight critical limitations, including difficulties in constraint satisfying, limited concept understanding, and the inherent probabilistic nature of outputs. We also propose key future directions, such as bridging the gap between generation and optimization. Although they are increasingly integrated in implementations, they differ fundamentally in both objectives and underlying mechanisms, necessitating a deeper understanding of their theoretical connections. Ultimately, this survey aims to provide a structured overview and a deeper insight into the strengths, limitations, and potential of GenAI in network optimization.

cs.NI↗

arXiv2026

AgriAgent: Contract-Driven Planning and Capability-Aware Tool Orchestration in Real-World Agriculture

Intelligent agent systems in real-world agricultural scenarios must handle diverse tasks under multimodal inputs, ranging from lightweight information understanding to complex multi-step execution. However, most existing approaches rely on a unified execution paradigm, which struggles to accommodate large variations in task complexity and incomplete tool availability commonly observed in agricultural environments. To address this challenge, we propose AgriAgent, a two-level agent framework for real-world agriculture. AgriAgent adopts a hierarchical execution strategy based on task complexity: simple tasks are handled through direct reasoning by modality-specific agents, while complex tasks trigger a contract-driven planning mechanism that formulates tasks as capability requirements and performs capability-aware tool orchestration and dynamic tool generation, enabling multi-step and verifiable execution with failure recovery. Experimental results show that AgriAgent achieves higher execution success rates and robustness on complex tasks compared to existing tool-centric agent baselines that rely on unified execution paradigms. All code, data will be released at after our work be accepted to promote reproducible research.

cs.CL↗

arXiv2026

MiMo-V2-Flash Technical Report

We present MiMo-V2-Flash, a Mixture-of-Experts (MoE) model with 309B total parameters and 15B active parameters, designed for fast, strong reasoning and agentic capabilities. MiMo-V2-Flash adopts a hybrid attention architecture that interleaves Sliding Window Attention (SWA) with global attention, with a 128-token sliding window under a 5:1 hybrid ratio. The model is pre-trained on 27 trillion tokens with Multi-Token Prediction (MTP), employing a native 32k context length and subsequently extended to 256k. To efficiently scale post-training compute, MiMo-V2-Flash introduces a novel Multi-Teacher On-Policy Distillation (MOPD) paradigm. In this framework, domain-specialized teachers (e.g., trained via large-scale reinforcement learning) provide dense and token-level reward, enabling the student model to perfectly master teacher expertise. MiMo-V2-Flash rivals top-tier open-weight models such as DeepSeek-V3.2 and Kimi-K2, despite using only 1/2 and 1/3 of their total parameters, respectively. During inference, by repurposing MTP as a draft model for speculative decoding, MiMo-V2-Flash achieves up to 3.6 acceptance length and 2.6x decoding speedup with three MTP layers. We open-source both the model weights and the three-layer MTP weights to foster open research and community collaboration.

cs.CL↗

arXiv2025

Hallucination Detection for LLM-based Text-to-SQL Generation via Two-Stage Metamorphic Testing

In Text-to-SQL generation, large language models (LLMs) have shown strong generalization and adaptability. However, LLMs sometimes generate hallucinations, i.e.,unrealistic or illogical content, which leads to incorrect SQL queries and negatively impacts downstream applications. Detecting these hallucinations is particularly challenging. Existing Text-to-SQL error detection methods, which are tailored for traditional deep learning models, face significant limitations when applied to LLMs. This is primarily due to the scarcity of ground-truth data. To address this challenge, we propose SQLHD, a novel hallucination detection method based on metamorphic testing (MT) that does not require standard answers. SQLHD splits the detection task into two sequentiial stages: schema-linking hallucination detection via eight structure-aware Metamorphic Relations (MRs) that perturb comparative words, entities, sentence structure or database schema, and logical-synthesis hallucination detection via nine logic-aware MRs that mutate prefix words, extremum expressions, comparison ranges or the entire database. In each stage the LLM is invoked separately to generate schema mappings or SQL artefacts; the follow-up outputs are cross-checked against their source counterparts through the corresponding MRs, and any violation is flagged as a hallucination without requiring ground-truth SQL. The experimental results demonstrate our method's superior performance in terms of the F1-score, which ranges from 69.36\% to 82.76\%. Additionally, SQLHD demonstrates superior performance over LLM Self-Evaluation methods, effectively identifying hallucinations in Text-to-SQL tasks.

cs.SE↗

arXiv2025

Energy-Efficient Omnidirectional Locomotion for Wheeled Quadrupeds via Predictive Energy-Aware Nominal Gait Selection

Wheeled-legged robots combine the efficiency of wheels with the versatility of legs, but face significant energy optimization challenges when navigating diverse environments. In this work, we present a hierarchical control framework that integrates predictive power modeling with residual reinforcement learning to optimize omnidirectional locomotion efficiency for wheeled quadrupedal robots. Our approach employs a novel power prediction network that forecasts energy consumption across different gait patterns over a 1-second horizon, enabling intelligent selection of the most energy-efficient nominal gait. A reinforcement learning policy then generates residual adjustments to this nominal gait, fine-tuning the robot's actions to balance energy efficiency with performance objectives. Comparative analysis shows our method reduces energy consumption by up to 35\% compared to fixed-gait approaches while maintaining comparable velocity tracking performance. We validate our framework through extensive simulations and real-world experiments on a modified Unitree Go1 platform, demonstrating robust performance even under external disturbances. Videos and implementation details are available at \href{https://sites.google.com/view/switching-wpg}{https://sites.google.com/view/switching-wpg}.

cs.RO↗

arXiv2025

Modeling and Experiments of an Injection-Locked Magnetron With Various Load Reflection Levels

In this article, we investigate the performance of an injection-locked 5.8-GHz continuous-wave magnetron with various load reflection levels. The load reflection is introduced to an equivalent magnetron model to theoretically evaluate the system performance. The effects of different load reflection levels on the magnetron's output are numerically analyzed. Experiments are performed while the load reflection is varied using an E-H tuner between a magnetron and a circulator. A narrower locking bandwidth is observed under constant injection power with increasing load reflection. The proper-mismatched system suppresses its sideband energy, thereby reducing phase noise. The experimental features qualitatively validate the theoretical analyses results. The investigation results also provide guidance for advanced applications in communication and high-energy physics based on injection-locked magnetrons.

physics.app-ph↗

arXiv2025

Enhancing Sample Efficiency and Exploration in Reinforcement Learning through the Integration of Diffusion Models and Proximal Policy Optimization

Proximal Policy Optimization (PPO) is widely used in continuous control due to its robustness and stable training, yet it remains sample-inefficient in tasks with expensive interactions and high-dimensional action spaces. This paper proposes PPO-DAP (PPO with Diffusion Action Prior), a strictly on-policy framework that improves exploration quality and learning efficiency without modifying the PPO objective. PPO-DAP follows a two-stage protocol. Offline, we pretrain a conditional diffusion action prior on logged trajectories to cover the action distribution supported by the behavior policy. Online, PPO updates the actor-critic only using newly collected on-policy rollouts, while the diffusion prior is adapted around the on-policy state distribution via parameter-efficient tuning (Adapter/LoRA) over a small parameter subset. For each on-policy state, the prior generates multiple action proposals and concentrates them toward high-value regions using critic-based energy reweighting and in-denoising gradient guidance. These proposals affect the actor only through a low-weight imitation loss and an optional soft KL regularizer to the prior; importantly, PPO gradients are never backpropagated through offline logs or purely synthetic trajectories. We further analyze the method from a dual-proximal perspective and derive a one-step performance lower bound. Across eight MuJoCo continuous-control tasks under a unified online budget of 1.0M environment steps, PPO-DAP consistently improves early learning efficiency (area under the learning curve over the first 40 epochs, ALC@40) and matches or exceeds the strongest on-policy baselines in final return on 6/8 tasks, with modest overhead (1.18+/-0.04x wall-clock time and 1.05+/-0.02x peak GPU memory relative to PPO).

cs.LG↗

arXiv2025

Quantum simulation of many-body dynamics with noise-robust Trotter decomposition based on symmetric structures

The Suzuki-Trotter decomposition, which digitalizes quantum time evolution, provides a promising framework for simulating quantum dynamics on quantum hardware and exploring quantum advantage over classical computation. However, conventional Trotter circuits require a large number of non-local gates, lowering their faithfulness to the ideal dynamics when implemented on current noisy quantum hardware. While most previous studies have focused on circuit optimization, we instead propose a new Trotter decomposition that is intrinsically circuit-efficient for simulating quantum dynamics on near-term devices. Our method substantially reduces both the residual error by Trotter decomposition and the number of CNOT operations compared to conventional Trotter decompositions by exploiting the symmetry of the target model to construct an effective Hamiltonian with fewer two-qubit gates. We demonstrate the noise robustness of the proposed approach through numerical simulations of a nine-site Heisenberg model under realistic noise, and further validate its experimental practicality on the IBM superconducting device, achieving a state fidelity exceeding $0.98$ when combined with quantum error mitigation in the three-site case. The proposed circuit design is also compatible with existing circuit optimization techniques. Our results establish a practical route toward noise-resilient quantum simulation in many-body dynamics.

quant-ph↗

arXiv2025

Dynamics of Anyon Clusters in Fractional Quantum Hall Fluids

In fractional quantum Hall fluids, the quasiparticle excitations are anyons with fractional charges and statistics. Effective interactions among the anyons can be induced by either model or realistic electron-electron (e-e) interactions. Without losing the generality, we investigate such phenomena for the Laughlin 1/3 and Moore-Read (MR) non-Abelian phases. Anyons display rich internal dynamics in both cases that will lead to interesting experimental consequences. In particular, bound states of two Laughlin anyons are preferred under short-range e-e interactions, leading to 2e/3 -- instead of e/3 -- effective charge carriers at low temperatures, which have been seen in several experiments. MR phases host two topologically distinct fusion channels: 1 and psi. The different effective interactions of e/4 anyons in the two sectors suggest the vanishing of the degeneracy of fusion channels when the e-e interaction is no longer its model Hamiltonian, in which case different bound states could also appear. This indicates the possibility of energetically manipulating the two types of anyons by tuning the bare e-e interactions. We point out that the results of recently developed high-resolution STM measurements will be affected by the effective anyon interactions, where anyons are clustered together after the tunneling of electrons. The low-lying parts of the local density of states affected by various anyon clusters are simulated for both Abelian and non-Abelian systems with (screened) Coulomb interactions.

cond-mat.str-el↗

arXiv2025

AgriGPT-Omni: A Unified Speech-Vision-Text Framework for Multilingual Agricultural Intelligence

Despite rapid advances in multimodal large language models, agricultural applications remain constrained by the lack of multilingual speech data, unified multimodal architectures, and comprehensive evaluation benchmarks. To address these challenges, we present AgriGPT-Omni, an agricultural omni-framework that integrates speech, vision, and text in a unified framework. First, we construct a scalable data synthesis and collection pipeline that converts agricultural texts and images into training data, resulting in the largest agricultural speech dataset to date, including 492K synthetic and 1.4K real speech samples across six languages. Second, based on this, we train the first agricultural omni-model via a three-stage paradigm: textual knowledge injection, progressive multimodal alignment, and GRPO-based reinforcement learning, enabling unified reasoning across languages and modalities. Third, we propose AgriBench-Omni-2K, the first tri-modal benchmark for agriculture, covering diverse speech-vision-text tasks and multilingual slices, with standardized protocols and reproducible tools. Experiments show that AgriGPT-Omni significantly outperforms general-purpose baselines on multilingual and multimodal reasoning as well as real-world speech understanding. All models, data, benchmarks, and code will be released to promote reproducible research, inclusive agricultural intelligence, and sustainable AI development for low-resource regions.

cs.CL↗

arXiv2025

A Robust Large-Period Discrete Time Crystal and its Signature in a Digital Quantum Computer

Discrete time crystals (DTCs) are novel out-of-equilibrium quantum states of matter which break time translational symmetry. DTCs have been extensively realized in experiments, particularly their subclass that is characterized by period-doubling dynamics due to its natural occurrence in a system of periodically driven two-level, e.g., spin-1/2, particles. The realization of DTCs beyond period-doubling, including their generalizations termed discrete quasicrystals has also been made in recent years, though such experiments typically involve higher spin particles. Constructing and observing DTCs beyond period-doubling in systems of two-level particles are generally still considered an open challenge due to the latter's $\mathbb{Z}_2$ symmetry that natively only leads to period-doubling. In this work, we developed an intuitive interacting system of two-level particles (qubits) that supports the more non-trivial period-quadrupling DTCs ($4T$-DTCs). Remarkably, by utilizing a variational algorithm, we are able to observe clear signatures of such $4T$-DTCs in a quantum processor despite the presence of considerable noise and the small number of available qubits. Our findings extend the landscape of time crystalline behavior by demonstrating a distinct realization of time crystallinity beyond standard period-doubling dynamics with qubits (two-level particles) on a NISQ-era digital quantum computer, as well as the potential of existing noisy intermediate-scale quantum devices for simulating exotic non-equilibrium quantum states of matter.

quant-ph↗

arXiv2025

AgriGPT-VL: Agricultural Vision-Language Understanding Suite

Despite rapid advances in multimodal large language models, agricultural applications remain constrained by the scarcity of domain-tailored models, curated vision-language corpora, and rigorous evaluation. To address these challenges, we present the AgriGPT-VL Suite, a unified multimodal framework for agriculture. Our contributions are threefold. First, we introduce Agri-3M-VL, the largest vision-language corpus for agriculture to our knowledge, curated by a scalable multi-agent data generator; it comprises 1M image-caption pairs, 2M image-grounded VQA pairs, 50K expert-level VQA instances, and 15K GRPO reinforcement learning samples. Second, we develop AgriGPT-VL, an agriculture-specialized vision-language model trained via a progressive curriculum of textual grounding, multimodal shallow/deep alignment, and GRPO refinement. This method achieves strong multimodal reasoning while preserving text-only capability. Third, we establish AgriBench-VL-4K, a compact yet challenging evaluation suite with open-ended and image-grounded questions, paired with multi-metric evaluation and an LLM-as-a-judge framework. Experiments show that AgriGPT-VL outperforms leading general-purpose VLMs on AgriBench-VL-4K, achieving higher pairwise win rates in the LLM-as-a-judge evaluation. Meanwhile, it remains competitive on the text-only AgriBench-13K with no noticeable degradation of language ability. Ablation studies further confirm consistent gains from our alignment and GRPO refinement stages. We will open source all of the resources to support reproducible research and deployment in low-resource agricultural settings.

cs.CL↗

arXiv2025

Machine Phenomenology: A Simple Equation Classifying Fast Radio Bursts

This work shows how human physical reasoning can guide machine-driven symbolic regression toward discovering empirical laws from observations. As an example, we derive a simple equation that classifies fast radio bursts (FRBs) into two distinct Gaussian distributions, indicating the existence of two physical classes. This human-AI workflow integrates feature selection, dimensional analysis, and symbolic regression: deep learning first analyzes CHIME Catalog 1 and identifies six independent parameters that collectively provide a complete description of FRBs; guided by Buckingham-$π$ analysis and correlation analysis, humans then construct dimensionless groups; finally, symbolic regression performed by the machine discovers the governing equation. When applied to the newer CHIME Catalog, the equation produces consistent results, demonstrating that it captures the underlying physics. This framework is applicable to a broad range of scientific domains.

astro-ph.IM↗

arXiv2025

Energy-Efficient Multi-UAV-Enabled MEC Systems over Space-Air-Ground Integrated Networks

With the development of artificial intelligence integrated next-generation communication networks, mobile users (MUs) are increasingly demanding the efficient processing of computation-intensive and latency-sensitive tasks. However, existing mobile computing networks struggle to support the rapidly growing computational needs of the MUs. Fortunately, space-air-ground integrated network (SAGIN) supported mobile edge computing (MEC) is regarded as an effective solution, offering the MUs multi-tier and efficient computing services. In this paper, we consider an SAGIN supported MEC system, where a low Earth orbit satellite and multiple unmanned aerial vehicles (UAVs) are dispatched to provide computing services for MUs. An energy efficiency maximization problem is formulated, with the joint optimization of the MU-UAV association, the UAV trajectory, the task offloading decision, the computing frequency, and the transmission power control. Since the problem is non-convex, we decompose it into four subproblems, and propose an alternating optimization based algorithm to solve it. Simulation results confirm that the proposed algorithm outperforms the benchmarks.

eess.SP↗

arXiv2025

SkyMoE: A Vision-Language Foundation Model for Enhancing Geospatial Interpretation with Mixture of Experts

The emergence of large vision-language models (VLMs) has significantly enhanced the efficiency and flexibility of geospatial interpretation. However, general-purpose VLMs remain suboptimal for remote sensing (RS) tasks. Existing geospatial VLMs typically adopt a unified modeling strategy and struggle to differentiate between task types and interpretation granularities, limiting their ability to balance local detail perception and global contextual understanding. In this paper, we present SkyMoE, a Mixture-of-Experts (MoE) vision-language model tailored for multimodal, multi-task RS interpretation. SkyMoE employs an adaptive router that generates task- and granularity-aware routing instructions, enabling specialized large language model experts to handle diverse sub-tasks. To further promote expert decoupling and granularity sensitivity, we introduce a context-disentangled augmentation strategy that creates contrastive pairs between local and global features, guiding experts toward level-specific representation learning. We also construct MGRS-Bench, a comprehensive benchmark covering multiple RS interpretation tasks and granularity levels, to evaluate generalization in complex scenarios. Extensive experiments on 21 public datasets demonstrate that SkyMoE achieves state-of-the-art performance across tasks, validating its adaptability, scalability, and superior multi-granularity understanding in remote sensing.

cs.CV↗

arXiv2025

Secure-Instruct: An Automated Pipeline for Synthesizing Instruction-Tuning Datasets Using LLMs for Secure Code Generation

Although Large Language Models (LLMs) show promising solutions to automated code generation, they often produce insecure code that threatens software security. Current approaches (e.g., SafeCoder) to improve secure code generation are limited by small, imbalanced instruction-tuning datasets. In this work, we present Secure-Instruct, a novel pipeline that automatically synthesizes high-quality vulnerable and secure code examples and instruction-tunes LLMs to align task description and secure code generation abilities. We evaluate Secure-Instruct on four representative LLMs using two security-related benchmarks: our own CWEBench and the existing CWEval. CWEBench comprises 93 scenarios on 44 CWEs, all without overlap with Secure-Instruct's synthetic instruction-tuning dataset, while CWEval covers 31 CWEs with 119 manually verified security-critical tasks. We find that Secure-Instruct improves both security and functional correctness in code generation. On CWEBench, Secure-Instruct substantially improves secure code generation, giving a 28.5% increase on average in secure ratio over the pre-trained models and outperforms SafeCoder by 12.6%. On CWEval, Secure-Instruct achieves an increase of 157.3% for CodeLlama-7B and 46.4% for Mistral-7B in Func-Sec@1 over pretrained models, and significantly outperforms SafeCoder.

cs.SE↗