arXiv subjects

Yujie Fu

Publications and source records attributed to Yujie Fu.

10 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

OmniaBench: Benchmarking General AI Agents Across Diverse Scenarios

Large language models are increasingly evolving from text generators into general agents capable of understanding user requests, invoking external tools, and completing complex tasks through interaction. However, existing agent benchmarks often focus on limited scenarios, tool ecosystems, or interaction formats, making it difficult to systematically characterize model capabilities across heterogeneous application settings. We introduce OmniaBench, a benchmark for evaluating general agents across diverse scenarios with explicit state spaces. We derive application-oriented scenario knowledge from app stores, product documents, industry resources, Web retrieval, and human refinement, forming a hierarchical taxonomy that spans ToC, ToB and ToE with 90 level-1 and 354 level-2 domains. Based on this taxonomy, we construct executable environments and synthesize single-turn and multi-turn tasks through four complementary routes: DAG, DAG-S, Solver, and Program. OmniaBench further introduces a ten-dimensional capability taxonomy and eight compositional atomic difficulty factors to support fine-grained evaluation and analysis. The resulting dataset contains 1,431 tasks, together with a challenging subset of 644 tasks designed to reduce evaluation cost and mitigate potential contamination of the full set after public release. The bench presents substantial challenges to current frontier models, with even Claude-Sonnet-5 and GPT-5.6-Sol achieving Overall Pass@1 scores of only 58.54 and 57.14, respectively. Further analyses reveal clear differences across domains and capabilities, as well as persistent limitations in planning, constraint maintenance, and adaptive correction. OmniaBench provides a broad and diagnostic benchmark for characterizing the capability boundaries of general agents.

cs.CL↗

arXiv2026

Data-driven Symbolic Closure for Turbulence Modeling in the Lattice Boltzmann Framework

Turbulence modeling within the Lattice Boltzmann Method (LBM) framework has long relied on traditional algebraic sub-grid scale (SGS) models, which often suffer from over-dissipation and lack of spatial selectivity near solid boundaries. In this work, we utilize Physical Symbolic Optimization (Phi-SO) to discover explicit analytical closures from high-fidelity DNS datasets of Taylor-Green Vortex (TGV) and Lid-Driven Cavity (LDC) flows. Central to our methodology is the integration of virtual dimensional analysis and non-linear tensor invariants, a strategy that enforces physical scaling laws directly within the symbolic search process. The resulting model exhibits a highly non-linear dependency on both strain-rate and rotation-rate invariants. Numerical validations confirm that this symbolic closure outperforms the standard Smagorinsky approach in capturing kinetic energy dissipation rate peaks and resolving delicate secondary corner vortices. Furthermore, the model exhibits robust zero-shot generalization to wall-bounded turbulent channel flow (Re_tau = 180) without the aid of any supplemental wall-damping corrections. This work highlights the potential of symbolic regression to uncover robust, interpretable physical laws for the next generation of intelligent computational fluid dynamics solvers.

physics.flu-dyn↗

arXiv2026

From inconsistency to decision: explainable operation and maintenance of battery energy storage systems

Battery Energy Storage Systems (BESSs) are increasingly critical to power-system stability, yet their operation and maintenance remain dominated by reactive, expert-dependent diagnostics. While cell-level inconsistencies provide early warning signals of degradation and safety risks, the lack of scalable and interpretable decision-support frameworks prevents these signals from being effectively translated into operational actions. Here we introduce an inconsistency-driven operation and maintenance paradigm for large-scale BESSs that systematically transforms routine monitoring data into explainable, decision-oriented guidance. The proposed framework integrates multi-dimensional inconsistency evaluation with large language model-based semantic reasoning to bridge the gap between quantitative diagnostics and practical maintenance decisions. Using eight months of field data from an in-service battery system comprising 3,564 cells, we demonstrate how electrical, thermal, and aging-related inconsistencies can be distilled into structured operational records and converted into actionable maintenance insights through a multi-agent framework. The proposed approach enables accurate and explainable responses to real-world operation and maintenance queries, reducing response time and operational cost by over 80% compared with conventional expert-driven practices. These results establish a scalable pathway for intelligent operation and maintenance of battery energy storage systems, with direct implications for reliability, safety, and cost-effective integration of energy storage into modern power systems.

eess.SY↗

arXiv2025

Carbon-based Microfabricated Organic Electrochemical Transistors Enabled by Printing and Laser Ablation

Organic electrochemical transistors (OECTs) are key bioelectronic devices, with applications in neuromorphics, sensing, and flexible electronics. However, their microfabrication typically relies on precious metal contacts manufactured via cleanroom processes. Here, we present a high-throughput additive-subtractive microfabrication strategy for metal-free, flexible OECTs using biodegradable materials and room-temperature processing. Additive manufacturing of large features is achieved via extrusion printing of a water-dispersed graphene ink to fabricate electrode contacts, and spin-coating of a cellulose acetate ink to form both the substrate and encapsulation layer. Combined with femtosecond laser ablation, this approach enables micrometer-resolution patterning of free-standing OECTs with channel openings down to 1 um and sheet resistance below 10 Ohm/sq. By tuning laser parameters, we demonstrate both selective and simultaneous ablation strategies, enabling the fabrication horizontal, vertical, and planar-gated OECTs, as well as complementary NOT gate inverters. Thermal degradation studies in air show that over 80% of the device mass decomposes below 360 deg C, providing a low-energy route for device disposal and addressing the environmental impact of electronic waste. This approach offers a cleanroom-free and lithography-free pathway toward the rapid prototyping of high-resolution, sustainable organic electronics, combining material circularity, process simplicity, and architectural versatility for next-generation bioelectronic applications.

physics.app-ph↗

arXiv2024

Efficient Quantization-Aware Training on Segment Anything Model in Medical Images and Its Deployment

Medical image segmentation is a critical component of clinical practice, and the state-of-the-art MedSAM model has significantly advanced this field. Nevertheless, critiques highlight that MedSAM demands substantial computational resources during inference. To address this issue, the CVPR 2024 MedSAM on Laptop Challenge was established to find an optimal balance between accuracy and processing speed. In this paper, we introduce a quantization-aware training pipeline designed to efficiently quantize the Segment Anything Model for medical images and deploy it using the OpenVINO inference engine. This pipeline optimizes both training time and disk storage. Our experimental results confirm that this approach considerably enhances processing speed over the baseline, while still achieving an acceptable accuracy level. The training script, inference script, and quantized model are publicly accessible at https://github.com/AVC2-UESTC/QMedSAM.

cs.CV↗

arXiv2024

Hazards in Daily Life? Enabling Robots to Proactively Detect and Resolve Anomalies

Existing household robots have made significant progress in performing routine tasks, such as cleaning floors or delivering objects. However, a key limitation of these robots is their inability to recognize potential problems or dangers in home environments. For example, a child may pick up and ingest medication that has fallen on the floor, posing a serious risk. We argue that household robots should proactively detect such hazards or anomalies within the home, and propose the task of anomaly scenario generation. We leverage foundational models instead of relying on manually labeled data to build simulated environments. Specifically, we introduce a multi-agent brainstorming approach, where agents collaborate and generate diverse scenarios covering household hazards, hygiene management, and child safety. These textual task descriptions are then integrated with designed 3D assets to simulate realistic environments. Within these constructed environments, the robotic agent learns the necessary skills to proactively discover and handle the proposed anomalies through task decomposition, and optimal learning approach selection. We demonstrate that our generated environment outperforms others in terms of task description and scene diversity, ultimately enabling robotic agents to better address potential household hazards.

cs.RO↗

arXiv2023

NPR: Nocturnal Place Recognition in Streets

Visual Place Recognition (VPR) is the task of retrieving database images similar to a query photo by comparing it to a large database of known images. In real-world applications, extreme illumination changes caused by query images taken at night pose a significant obstacle that VPR needs to overcome. However, a training set with day-night correspondence for city-scale, street-level VPR does not exist. To address this challenge, we propose a novel pipeline that divides VPR and conquers Nocturnal Place Recognition (NPR). Specifically, we first established a street-level day-night dataset, NightStreet, and used it to train an unpaired image-to-image translation model. Then we used this model to process existing large-scale VPR datasets to generate the VPR-Night datasets and demonstrated how to combine them with two popular VPR pipelines. Finally, we proposed a divide-and-conquer VPR framework and provided explanations at the theoretical, experimental, and application levels. Under our framework, previous methods can significantly improve performance on two public datasets, including the top-ranked method.

cs.CV↗

arXiv2022

Stake-governed tug-of-war and the biased infinity Laplacian

In tug-of-war, two players compete by moving a counter along edges of a graph, each winning the right to move at a given turn according to the flip of a possibly biased coin. The game ends when the counter reaches the boundary, a fixed subset of the vertices, at which point one player pays the other an amount determined by the boundary vertex. Economists and mathematicians have independently studied tug-of-war for many years, focussing respectively on resource-allocation forms of the game, in which players iteratively spend precious budgets in an effort to influence the bias of the coins that determine the turn victors; and on PDE arising in fine mesh limits of the constant-bias game in a Euclidean setting. In this article, we offer a mathematical treatment of a class of tug-of-war games with allocated budgets: each player is initially given a fixed budget which she draws on throughout the game to offer a stake at the start of each turn, and her probability of winning the turn is the ratio of her stake and the sum of the two stakes. We consider the game played on a tree, with boundary being the set of leaves, and the payment function being the indicator of a single distinguished leaf. We find the game value and the essentially unique Nash equilibrium of a leisurely version of the game, in which the move at any given turn is cancelled with constant probability after stakes have been placed. We show that the ratio of the players' remaining budgets is maintained at its initial value $\lambda$; game value is a biased infinity harmonic function; and the proportion of remaining budget that players stake at a given turn is given in terms of the spatial gradient and the $\lambda$-derivative of game value. We also indicate examples in which the solution takes a different form in the non-leisurely game.

math.PR↗

arXiv2021

Scale-Net: Learning to Reduce Scale Differences for Large-Scale Invariant Image Matching

Most image matching methods perform poorly when encountering large scale changes in images. To solve this problem, firstly, we propose a scale-difference-aware image matching method (SDAIM) that reduces image scale differences before local feature extraction, via resizing both images of an image pair according to an estimated scale ratio. Secondly, in order to accurately estimate the scale ratio, we propose a covisibility-attention-reinforced matching module (CVARM) and then design a novel neural network, termed as Scale-Net, based on CVARM. The proposed CVARM can lay more stress on covisible areas within the image pair and suppress the distraction from those areas visible in only one image. Quantitative and qualitative experiments confirm that the proposed Scale-Net has higher scale ratio estimation accuracy and much better generalization ability compared with all the existing scale ratio estimation methods. Further experiments on image matching and relative pose estimation tasks demonstrate that our SDAIM and Scale-Net are able to greatly boost the performance of representative local features and state-of-the-art local feature matching methods.

cs.CV↗