Searcharxiv⌕ Search

arXiv subjects

Shaowei Chen

Publications and source records attributed to Shaowei Chen.

18 recordsLinked to original sources
arXiv2026

Iris: Climbing to the Search Frontier

We present Iris-mini and Iris-pro, two search agents trained at the 35B-A3B and 397B-A17B scales, together with the data pipeline and training recipe behind them. Tasks are reverse-constructed from the hyperlink structure of a web corpus: we author multi-hop chains over an entity graph distilled from a seed page and its out-links, rewrite every non-answer entity into a descriptive reference so that no clue can be resolved by string matching, and admit only questions that a reference model fails closed-book yet solves once the supporting evidence is supplied. These questions are then turned into trajectories, which are filtered at both the trajectory and the turn level before SFT. The policy is then optimized by RL against live search, with the reward judge and the observation summarizer served inside the training cluster, and with over-long rollouts interrupted at the request level and resumed from their committed prefix at the next step. We alternate the two stages in a procedure we call SFT-RL climbing, returning the hardest solved and most efficient rollouts of each RL round to the next supervised pass. Because inference-time context management is worth more on these benchmarks than most reported differences between systems, we evaluate every benchmark both with and without it, holding the tool set, the context limit, and the judge fixed. All results come from a single ReAct agent, with no sub-agents and no test-time verification. With management enabled, on BrowseComp, BrowseComp-ZH, DeepSearchQA, and HLE the two models reach $82.2/84.8/86.9/52.3$ and $88.6/85.1/92.9/56.4$, the strongest overall results among open-source search agents in their respective parameter ranges. We plan to release the model weights together with the complete recipe for data construction, training, and evaluation.

cs.AI↗
arXiv2025

Data-Efficient Motor Condition Monitoring with Time Series Foundation Models

Motor condition monitoring is essential for ensuring system reliability and preventing catastrophic failures. However, data-driven diagnostic methods often suffer from sparse fault labels and severe class imbalance, which limit their effectiveness in real-world applications. This paper proposes a motor condition monitoring framework that leverages the general features learned during pre-training of two time series foundation models, MOMENT and Mantis, to address these challenges. By transferring broad temporal representations from large-scale pre-training, the proposed approach significantly reduces dependence on labeled data while maintaining high diagnostic accuracy. Experimental results show that MOMENT achieves nearly twice the performance of conventional deep learning models using only 1% of the training data, whereas Mantis surpasses state-of-the-art baselines by 22%, reaching 90% accuracy with the same data ratio. These results demonstrate the strong generalization and data efficiency of time series foundation models in fault diagnosis, providing new insights into scalable and adaptive frameworks for intelligent motor condition monitoring.

eess.SP↗
arXiv2025

Parallelizable Complex Neural Dynamics Models for PMSM Temperature Estimation with Hardware Acceleration

Accurate and efficient thermal dynamics models of permanent magnet synchronous motors are vital to efficient thermal management strategies. Physics-informed methods combine model-based and data-driven methods, offering greater flexibility than model-based methods and superior explainability compared to data-driven methods. Nonetheless, there are still challenges in balancing real-time performance, estimation accuracy, and explainability. This paper presents a hardware-efficient complex neural dynamics model achieved through the linear decoupling, diagonalization, and reparameterization of the state-space model, introducing a novel paradigm for the physics-informed method that offers high explainability and accuracy in electric motor temperature estimation tasks. We validate this physics-informed method on an NVIDIA A800 GPU using the JAX machine learning framework, parallel prefix sum algorithm, and Compute Unified Device Architecture (CUDA) platform. We demonstrate its superior estimation accuracy and parallelizable hardware acceleration capabilities through experimental evaluation on a real electric motor.

eess.SY↗
arXiv2025

Tongyi DeepResearch Technical Report

We present Tongyi DeepResearch, an agentic large language model, which is specifically designed for long-horizon, deep information-seeking research tasks. To incentivize autonomous deep research agency, Tongyi DeepResearch is developed through an end-to-end training framework that combines agentic mid-training and agentic post-training, enabling scalable reasoning and information seeking across complex tasks. We design a highly scalable data synthesis pipeline that is fully automatic, without relying on costly human annotation, and empowers all training stages. By constructing customized environments for each stage, our system enables stable and consistent interactions throughout. Tongyi DeepResearch, featuring 30.5 billion total parameters, with only 3.3 billion activated per token, achieves state-of-the-art performance across a range of agentic deep research benchmarks, including Humanity's Last Exam, BrowseComp, BrowseComp-ZH, WebWalkerQA, xbench-DeepSearch, FRAMES and xbench-DeepSearch-2510. We open-source the model, framework, and complete solutions to empower the community.

cs.CL↗
arXiv2025

Task-KV: Task-aware KV Cache Optimization via Semantic Differentiation of Attention Heads

KV cache is a widely used acceleration technique for large language models (LLMs) inference. However, its memory requirement grows rapidly with input length. Previous studies have reduced the size of KV cache by either removing the same number of unimportant tokens for all attention heads or by allocating differentiated KV cache budgets for pre-identified attention heads. However, due to the importance of attention heads varies across different tasks, the pre-identified attention heads fail to adapt effectively to various downstream tasks. To address this issue, we propose Task-KV, a method that leverages the semantic differentiation of attention heads to allocate differentiated KV cache budgets across various tasks. We demonstrate that attention heads far from the semantic center (called heterogeneous heads) make an significant contribution to task outputs and semantic understanding. In contrast, other attention heads play the role of aggregating important information and focusing reasoning. Task-KV allocates full KV cache budget to heterogeneous heads to preserve comprehensive semantic information, while reserving a small number of recent tokens and attention sinks for non-heterogeneous heads. Furthermore, we innovatively introduce middle activations to preserve key contextual information aggregated from non-heterogeneous heads. To dynamically perceive semantic differences among attention heads, we design a semantic separator to distinguish heterogeneous heads from non-heterogeneous ones based on their distances from the semantic center. Experimental results on multiple benchmarks and different model architectures demonstrate that Task-KV significantly outperforms existing baseline methods.

cs.CL↗
arXiv2023

Physics-Informed Neural Networks for Prognostics and Health Management of Lithium-Ion Batteries

For Prognostics and Health Management (PHM) of Lithium-ion (Li-ion) batteries, many models have been established to characterize their degradation process. The existing empirical or physical models can reveal important information regarding the degradation dynamics. However, there are no general and flexible methods to fuse the information represented by those models. Physics-Informed Neural Network (PINN) is an efficient tool to fuse empirical or physical dynamic models with data-driven models. To take full advantage of various information sources, we propose a model fusion scheme based on PINN. It is implemented by developing a semi-empirical semi-physical Partial Differential Equation (PDE) to model the degradation dynamics of Li-ion batteries. When there is little prior knowledge about the dynamics, we leverage the data-driven Deep Hidden Physics Model (DeepHPM) to discover the underlying governing dynamic models. The uncovered dynamics information is then fused with that mined by the surrogate neural network in the PINN framework. Moreover, an uncertainty-based adaptive weighting method is employed to balance the multiple learning tasks when training the PINN. The proposed methods are verified on a public dataset of Li-ion Phosphate (LFP)/graphite batteries.

eess.SP↗
arXiv2021

Higher topological type semiclassical states for fractional nonlinear elliptic equations

In this paper, we are concerned with semiclassical states to the following fractional nonlinear elliptic equation, \begin{align*} \eps^{2s}(-\Delta)^s u + V(x) u=\mathcal{N}(|u|)u \quad \mbox{in} \,\,\, \R^N, \end{align*} where $0 0$ is a small parameter, $N>2s$, $V \in C^1(\R^N, \R^+)$ and $\mathcal{N}\in C(\R, \R)$. The nonlinearity has Sobolev subcritical, critical or supercritical growth. The fractional Laplacian $(-\Delta)^s$ is characterized as $\mathcal{F}((-\Delta)^{s}u)(\xi)=|\xi|^{2s} \mathcal{F}(u)(\xi)$ for $\xi \in \R^N$, where $\mathcal{F}$ denotes the Fourier transform. We construct positive semiclassical states and an infinite sequence of sign-changing semiclassical states with higher energies clustering near the local minimum points of the potential $V$. The solutions are of higher topological type, which are obtained from a minimax characterization of higher dimensional symmetric linking structure via the symmetric mountain pass theorem. They correspond to critical points of the underlying energy functional at energy levels where compactness condition breaks down. The proofs are mainly based on penalization methods, s-harmonic extension theories and blow-up arguments along with local type Pohozaev identities.

math.AP↗
arXiv2021

Higher topological type semiclassical states for Sobolev critical Dirac equations with degenerate potential

In this paper, we are concerned with semiclassical states to the following Sobolev critical Dirac equation with degenerate potential, \begin{align*} -\textnormal{i} \eps \alpha \cdot \nabla u + a \beta u + V(x) u=|u|^{q-2} u + |u| u \quad \mbox{in} \,\, \R^3, \end{align*} where $u:\mathbb{R}^3\rightarrow \mathbb{C}^4$, $2 0$ is a small parameter, $a>0$ is a constant, $\alpha=(\alpha_1, \alpha_2, \alpha_3)$, $\alpha_j$ and $\beta$ are $4 \times 4$ Pauli-Dirac matrices. We construct an infinite sequence of higher topological type semiclassical states with higher energies concentrating around the local minimum points of the degenerate potential $V$. The solutions are obtained from a minimax characterization of higher dimensional symmetric linking structure, which correspond to critical points of the underlying energy functional at energy levels where compactness condition breaks down. Our approach is variational, which mainly relies on penalization method and blow-up arguments along with local type Pohozaev identity.

math.AP↗
arXiv2021

Bidirectional Machine Reading Comprehension for Aspect Sentiment Triplet Extraction

Aspect sentiment triplet extraction (ASTE), which aims to identify aspects from review sentences along with their corresponding opinion expressions and sentiments, is an emerging task in fine-grained opinion mining. Since ASTE consists of multiple subtasks, including opinion entity extraction, relation detection, and sentiment classification, it is critical and challenging to appropriately capture and utilize the associations among them. In this paper, we transform ASTE task into a multi-turn machine reading comprehension (MTMRC) task and propose a bidirectional MRC (BMRC) framework to address this challenge. Specifically, we devise three types of queries, including non-restrictive extraction queries, restrictive extraction queries and sentiment classification queries, to build the associations among different subtasks. Furthermore, considering that an aspect sentiment triplet can derive from either an aspect or an opinion expression, we design a bidirectional MRC structure. One direction sequentially recognizes aspects, opinion expressions, and sentiments to obtain triplets, while the other direction identifies opinion expressions first, then aspects, and at last sentiments. By making the two directions complement each other, our framework can identify triplets more comprehensively. To verify the effectiveness of our approach, we conduct extensive experiments on four benchmark datasets. The experimental results demonstrate that BMRC achieves state-of-the-art performances.

cs.CL↗
arXiv2014

Existence of a nontrivial solution for a strongly indefinite periodic Schrodinger-Poisson system

We consider the Schr\"odinger-Poisson system \begin{eqnarray}\left\{\begin{array} [c]{ll} -\Delta u+V(x) u+|u|^{p-2}u=\lambda \phi u, & \mbox{in}\mathbb{R}^{3},\\ -\Delta\phi= u^{2}, & \mbox{in}\mathbb{R}^{3}. \end{array} \right.\nonumber \end{eqnarray} where $\lambda>0$ is a parameter, $3< p<6$, $V\in C(\mathbb{R}^{3}) $ is $1$-periodic in $x_j$ for $j = 1,2,3$ and 0 is in a spectral gap of the operator $-\Delta+V$. This system is strongly indefinite, i.e., the operator $-\Delta+V$ has infinite-dimensional negative and positive spaces and it has a competitive interplay of the nonlinearities $|u|^{p-2}u$ and $\lambda \phi u$. Moreover, the functional corresponding to this system does not satisfy the Palai-Smale condition. Using a new infinite-dimensional linking theorem, we prove that, for sufficiently small $\lambda>0,$ this system has a nontrivial solution.

math.AP↗
arXiv2014

Existence of nontrivial solutions for periodic Schrodinger equations with new nonlinearities

We study the Schr\"{o}dinger equation: \begin{eqnarray} - \Delta u+V(x)u+f(x,u)=0,\qquad u\in H^{1}(\mathbb{R}^{N}),\nonumber \end{eqnarray} where $V$ is periodic and $f$ is periodic in the $x$-variables, $0$ is in a gap of the spectrum of the operator $-\Delta+V$. We prove that under some new assumptions for $f$, this equation has a nontrivial solution. Our assumptions for the nonlinearity $f$ are very weak and greatly different from the known assumptions in the literature.

math.AP↗
arXiv2013

Existence of nontrivial solutions for asymptotically linear periodic Schr\"odinger equations

We study the Schr\"{o}dinger equation: \begin{equation} - \Delta u+V(x)u=f(x,u) ,\qquad u\in H^{1}(\mathbb{R}^{N}),\nonumber \end{equation} where $V$ is periodic and $f$ is periodic in the $x$-variables, 0 is in a gap of the spectrum of the operator $-\Delta+V$ and $f$ is asymptotically linear as $|u|\rightarrow+\infty.$ We prove that under some asymptotically linear assumptions for $f$, this equation has a nontrivial solution. Our assumptions for $f$ are different from the classical assumptions raised by G. B. Li and A. Szulkin in

math.AP↗
arXiv2012

Solution manifolds for some semilinear Schrodinger equations

In this paper, we study the following semilinear Schr\"odinger system $$ -\triangle u+u=(1+K_\alpha(\epsilon x))|u|^{p-2}u\ in \mathbb{R}^N, u\in H^1(\mathbb{R}^N) $$ where $3\leq p<2^*$ and $\epsilon>0$, $\alpha>0$ are small parameters. Under some conditions on $K_\alpha$ and the parameters $\alpha$ and $\epsilon$, we show that this equation exist solution manifolds.

math.AP↗
arXiv2012

Existence and concentration of semiclassical states for nonlinear Schrodinger equations

In this paper, we study the following semilinear Schr\"odinger equation $$ -\epsilon^2\triangle u+ u+ V(x)u=f(u),\ u\in H^{1}(\mathbb{R}^{N}), $$ where $N\geq 2$ and $\epsilon>0$ is a small parameter. The function $V$ is bounded in $\mathbb{R}^N$, $\inf_{\mathbb{R}^N}(1+V(x))>0$ and it has a possibly degenerate isolated critical point. Under some conditions on $f,$ we prove that as $\epsilon\rightarrow 0,$ this equation has a solution which concentrates at the critical point of $V$.}

math.AP↗
arXiv2009

Results on entire solutions for a degenerate critical elliptic equation with anisotropic coefficients

In this paper, we study the following degenerate critical elliptic equations with anisotropic coefficients $$ -div(|x_{N}|^{2\alpha}\nabla u)=K(x)|x_{N}|^{\alpha\cdot 2^{*}(s)-s}|u|^{2^{*}(s)-2}u {in} \mathbb{R}^{N} $$ where $x=(x_{1},...,x_{N})\in\mathbb{R}^{N},$ $N\geq 3,$ $\alpha>1/2,$ $0\leq s\leq 2$ and $2^{*}(s)=2(N-s)/(N-2).$ Some basic properties of the degenerate elliptic operator $-div(|x_{N}|^{2\alpha}\nabla u)$ are investigated and some regularity, symmetry and uniqueness results for entire solutions of this equation are obtained. We also get some variational identities for solutions of this equation. As a consequence, we obtain some nonexistence results for solutions of this equation.

math.AP↗
arXiv2008

Multi-bump Solutions for a Strongly Indefinite Semilinear Schr\"odinger Equation Without Symmetry or convexity Assumptions

In this paper, we study the following semilinear Schr\"odinger equation with periodic coefficient: $$-\triangle u +V(x)u=f(x,u), u\in H^{1}(\mathbb{R}^{N}).$$ The functional corresponding to this equation possesses strongly indefinite structure. The nonlinear term $f(x,t)$ satisfies some superlinear growth conditions and need not be odd or increasing strictly in $t$. Using a new variational reduction method and a generalized Morse theory, we proved that this equation has infinitely many geometrically different solutions. Furthermore, if the solutions of this equation under some energy level are isolated, then we can show that this equation has infinitely many $m-$bump solutions for any positive integer $m\geq 2.$

math.AP↗