arXiv subjects

Lu Chen

Publications and source records attributed to Lu Chen.

At least 109 records · Page 6Linked to original sources

arXiv2025

Exploring Efficiency Frontiers of Thinking Budget in Medical Reasoning: Scaling Laws between Computational Resources and Reasoning Quality

This study presents the first comprehensive evaluation of thinking budget mechanisms in medical reasoning tasks, revealing fundamental scaling laws between computational resources and reasoning quality. We systematically evaluated two major model families, Qwen3 (1.7B to 235B parameters) and DeepSeek-R1 (1.5B to 70B parameters), across 15 medical datasets spanning diverse specialties and difficulty levels. Through controlled experiments with thinking budgets ranging from zero to unlimited tokens, we establish logarithmic scaling relationships where accuracy improvements follow a predictable pattern with both thinking budget and model size. Our findings identify three distinct efficiency regimes: high-efficiency (0 to 256 tokens) suitable for real-time applications, balanced (256 to 512 tokens) offering optimal cost-performance tradeoffs for routine clinical support, and high-accuracy (above 512 tokens) justified only for critical diagnostic tasks. Notably, smaller models demonstrate disproportionately larger benefits from extended thinking, with 15 to 20% improvements compared to 5 to 10% for larger models, suggesting a complementary relationship where thinking budget provides greater relative benefits for capacity-constrained models. Domain-specific patterns emerge clearly, with neurology and gastroenterology requiring significantly deeper reasoning processes than cardiovascular or respiratory medicine. The consistency between Qwen3 native thinking budget API and our proposed truncation method for DeepSeek-R1 validates the generalizability of thinking budget concepts across architectures. These results establish thinking budget control as a critical mechanism for optimizing medical AI systems, enabling dynamic resource allocation aligned with clinical needs while maintaining the transparency essential for healthcare deployment.

cs.CL↗

arXiv2025

47B Mixture-of-Experts Beats 671B Dense Models on Chinese Medical Examinations

The rapid advancement of large language models(LLMs) has prompted significant interest in their potential applications in medical domains. This paper presents a comprehensive benchmark evaluation of 27 state-of-the-art LLMs on Chinese medical examination questions, encompassing seven medical specialties across two professional levels. We introduce a robust evaluation framework that assesses model performance on 2,800 carefully curated questions from cardiovascular, gastroenterology, hematology, infectious diseases, nephrology, neurology, and respiratory medicine domains. Our dataset distinguishes between attending physician and senior physician difficulty levels, providing nuanced insights into model capabilities across varying complexity. Our empirical analysis reveals substantial performance variations among models, with Mixtral-8x7B achieving the highest overall accuracy of 74.25%, followed by DeepSeek-R1-671B at 64.07%. Notably, we observe no consistent correlation between model size and performance, as evidenced by the strong performance of smaller mixture-of-experts architectures. The evaluation demonstrates significant performance gaps between medical specialties, with models generally performing better on cardiovascular and neurology questions compared to gastroenterology and nephrology domains. Furthermore, our analysis indicates minimal performance degradation between attending and senior physician levels for top-performing models, suggesting robust generalization capabilities. This benchmark provides critical insights for the deployment of LLMs in medical education and clinical decision support systems, highlighting both the promise and current limitations of these technologies in specialized medical contexts.

cs.CL↗

arXiv2025

Is GPT-OSS Good? A Comprehensive Evaluation of OpenAI's Latest Open Source Models

In August 2025, OpenAI released GPT-OSS models, its first open weight large language models since GPT-2 in 2019, comprising two mixture of experts architectures with 120B and 20B parameters. We evaluated both variants against six contemporary open source large language models ranging from 14.7B to 235B parameters, representing both dense and sparse designs, across ten benchmarks covering general knowledge, mathematical reasoning, code generation, multilingual understanding, and conversational ability. All models were tested in unquantised form under standardised inference settings, with statistical validation using McNemars test and effect size analysis. Results show that gpt-oss-20B consistently outperforms gpt-oss-120B on several benchmarks, such as HumanEval and MMLU, despite requiring substantially less memory and energy per response. Both models demonstrate mid-tier overall performance within the current open source landscape, with relative strength in code generation and notable weaknesses in multilingual tasks. These findings provide empirical evidence that scaling in sparse architectures may not yield proportional performance gains, underscoring the need for further investigation into optimisation strategies and informing more efficient model selection for future open source deployments. More details and evaluation scripts are available at https://ai-agent-lab.github.io/gpt-oss (Project Webpage).

cs.CL↗

arXiv2025

Classification of positive solutions of Hardy-Sobolev equation without the finite volume constraints

In this paper, we are concerned with the critical Hardy-Sobolev equation \begin{equation*} -Δ_{p}u = \frac{u^{p^{*}_s-1}}{|x|^{s}}, \ \ x\in \mathbb{R}^n \end{equation*} where $p^{*}_s = \frac{(n-s)p}{n-p}$ denotes the critical Hardy-Sobolev exponent. We classify the positive solutions of this equation for $0 < s < \frac{p-1}{p}$ and $\frac{(2s+n+1)+\sqrt{(2s+n+1)^2-12s}}{6} \leq p < n$ without finite volume constraints, which extends Ou's result in \cite{9} in the literature. The method is based on constructing suitable vector fields integral inequality and using Newton's type inequality.

math.AP↗

arXiv2025

Classification of positive solutions of critical anisotropic Sobolev equation without the finite volume constraint

In this paper, we classify all positive solutions of the critical anisotropic Sobolev equation \begin{equation}\label{0.1} -Δ^{H}_{p}u = u^{p^{*}-1}, \ \ x\in \mathbb{R}^n \end{equation} without the finite volume constraint for $n \geq 3$ and $p_n(Λ) < p < n$, where $p^{*} = \frac{np}{n-p}$ denotes the critical Sobolev exponent, $-Δ^{H}_{p}=-div(H^{p-1}(\cdot)\nabla H(\cdot))$ denotes the anisotropic $p$-Laplace operator and $Λ= λ\max\limits_{\substack{ξ\in \mathbb{R}^n\\1 \leq i, j \leq n}}\left\{\frac{|ξ|^{2}(\nabla^{2}_{ij}H^{p}(ξ))} {p(p-1)H^{p}(ξ)}\right\}$. By employing a novel approach based on invariant tensors technique, and using a Kato-type inequality, we prove that the positive solutions of \eqref{0.1} can be classified for $p_n(Λ) \leq p < n$, where $p_n(Λ)$ depends explicitly on $Λ$. This result removes the finite volume assumption on the classification of critical anisotropic $p$-Laplace equation which was obtained by Ciraolo-Figalli-Roncoroni in the literature \cite{CFR}. In particular, this results capture the precise dependence of critical exponents $p$ on both $n$ and $Λ$.

math.AP↗

arXiv2025

FuncGenFoil: Airfoil Generation and Editing Model in Function Space

Aircraft manufacturing is the jewel in the crown of industry, in which generating high-fidelity airfoil geometries with controllable and editable representations remains a fundamental challenge. Existing deep learning methods, which typically rely on predefined parametric representations (e.g., Bézier) or discrete point sets, face an inherent trade-off between expressive power and resolution adaptability. To tackle this challenge, we introduce FuncGenFoil, a novel function-space generative model that directly reconstructs airfoil geometries as function curves. Our method inherits the advantages of arbitrary-resolution sampling and smoothness from parametric functions, as well as the strong expressiveness of discrete point-based representations. Empirical evaluations demonstrate that FuncGenFoil improves upon state-of-the-art methods in airfoil generation, achieving a relative 74.4% reduction in label error and a 23.2% increase in diversity on the AF-200K dataset. Our results highlight the advantages of function-space modeling for aerodynamic shape optimization, offering a powerful and flexible framework for high-fidelity airfoil design.

cs.LG↗

arXiv2025

Nex-N1: Agentic Models Trained via a Unified Ecosystem for Large-Scale Environment Construction

The evolution of Large Language Models (LLMs) from passive responders to autonomous agents necessitates a fundamental shift in learning paradigms -- from static imitation to incentive-driven decision making. However, this transition is significantly impeded by the lack of scalable infrastructure capable of constructing high-quality interaction signals for effective policy learning. To address this, we introduce a comprehensive method designed to systematically scale the diversity and complexity of interactive environments. Our method realizes this scaling by addressing three orthogonal dimensions: (1) Complexity: NexAU, a flexible agent framework that supports building complex agent hierarchies via simple configurations; (2) Diversity: NexA4A automatically generates diverse agent hierarchies from natural language to cover infinite domains; and (3) Fidelity: NexGAP bridges the simulation-reality gap by integrating dynamic real-world environment for grounded trajectories synthesis. We train Nex-N1 upon the diverse and complex interactive environments established by our infrastructure. Empirical results on benchmarks such as SWE-bench and tau2 demonstrate that Nex-N1 consistently outperforms SOTA open-source models and achieves competitive performance against frontier proprietary models on complex agentic tasks. We open-source the Nex ecosystem and model weights to facilitate further research.

cs.CL↗

arXiv2025

Think Fast: Real-Time Kinodynamic Belief-Space Planning for Projectile Interception

Intercepting fast moving objects, by its very nature, is challenging because of its tight time constraints. This problem becomes further complicated in the presence of sensor noise because noisy sensors provide, at best, incomplete information, which results in a distribution over target states to be intercepted. Since time is of the essence, to hit the target, the planner must begin directing the interceptor, in this case a robot arm, while still receiving information. We introduce an tree-like structure, which is grown using kinodynamic motion primitives in state-time space. This tree-like structure encodes reachability to multiple goals from a single origin, while enabling real-time value updates as the target belief evolves and seamless transitions between goals. We evaluate our framework on an interception task on a 6 DOF industrial arm (ABB IRB-1600) with an onboard stereo camera (ZED 2i). A robust Innovation-based Adaptive Estimation Adaptive Kalman Filter (RIAE-AKF) is used to track the target and perform belief updates.

cs.RO↗

arXiv2025

Beyond Relational: Semantic-Aware Multi-Modal Analytics with LLM-Native Query Optimization

Multi-modal analytical processing has the potential to transform applications in e-commerce, healthcare, entertainment, and beyond. However, real-world adoption remains elusive due to the limited ability of traditional relational query operators to capture query semantics. The emergence of foundation models, particularly the large language models (LLMs), opens up new opportunities to develop flexible, semantic-aware data analytics systems that transcend the relational paradigm. We present Nirvana, a multi-modal data analytics framework that incorporates programmable semantic operators while leveraging both logical and physical query optimization strategies, tailored for LLM-driven semantic query processing. Nirvana addresses two key challenges. First, it features an agentic logical optimizer that uses natural language-specified transformation rules and random-walk-based search to explore vast spaces of semantically equivalent query plans -- far beyond the capabilities of conventional optimizers. Second, it introduces a cost-aware physical optimizer that selects the most effective LLM backend for each operator using a novel improvement-score metric. To further enhance efficiency, Nirvana incorporates computation reuse and evaluation pushdown techniques guided by model capability hypotheses. Experimental evaluations on three real-world benchmarks demonstrate that Nirvana is able to reduce end-to-end runtime by 10%--85% and reduces system processing costs by 76% on average, outperforming state-of-the-art systems at both efficiency and scalability.

cs.DB↗

arXiv2025

Implications of the Two-Component Dark Energy Model for Hubble Tension

Dark energy plays a crucial role in the evolution of cosmic expansion. In most studies, dark energy is considered a single dynamic component. In fact, multi-component dark energy models may theoretically explain the accelerated expansion of the universe as well. In our previous research, we constructed the $w_{\rm{n}}$CDM ($n=2, 3, 5$) models and conducted numerical research, finding strong observational support when the value of n is small. Based on our results, both the $χ^2$ and Akaike information criterion (AIC) favor the $w_{\rm{2}}$CDM model more than the $w_0w_{\rm{a}}$CDM model. However, previous studies were limited to two equal-component dark energy models, failing to consider the component proportions as variables. Therefore, we will further explore the $w_{\rm{2}}$CDM model. To simplify the model, we fix $w = -1$ in one component and set the other component to $w_{\rm{de2}}$, varying the proportions of both components in the population. Under different $w_{\rm{de2}}$, we obtain the one-dimensional distribution of ${H}_{0}$ with respect to $f_{\rm{de2}}$. Further fitting reveals the evolution of ${H}_{0}$ under varying $w_{\rm{de2}}$ and $f_{\rm{de2}}$. We also perform the same operation on $χ^2$. To evaluate the error of fitting, we introduce two indicators, $\text{R}^{2}_{\text{adj}}$ and MAPE, to quantify the fitting ability of our models. We find that when $w_{\rm{de2}}$ is less than -1, ${H}_{0}$ increases with the decrease of $w_{\rm{de2}}$ and the increase of $f_{\rm{de2}}$, effectively alleviating ${H}_{0}$ tension. For $χ^2$, it still prefers the $Λ$CDM model, and the $w_{\rm{2}}$CDM model will decrease significantly when it approaches the $Λ$CDM model. The excellent performance of $\text{R}^{2}_{\text{adj}}$ and MAPE further proves that our model has an outstanding fitting effect and extremely high reliability.

astro-ph.CO↗

arXiv2025

Competition between Weak Localization and Antilocalization of Dirac-like Fermions in a Spin-Polarized Two-Dimensional Electron Gas at KTaO3 (111) Interface

Quantum transport phenomena in two-dimensional electron gases (2DEGs) at oxide interfaces have garnered significant interest owing to their potential in spintronic and quantum information technologies. Here, we systematically investigate the quantum conductance corrections of spin-polarized 2DEGs formed at the interfaces between two insulating oxides, ferromagnetic EuTiO3 (ETO) films and (111)-oriented KTaO3 (KTO) substrates. The anomalous Hall effect and hysteretic magnetoresistance provide clear evidence for long-range ferromagnetic order in the 2DEGs, which could be attributed to interfacial Eu doping in combination with the magnetic proximity effect of the ETO layer. The breaking of time-reversal symmetry by ferromagnetism in the 2DEGs, and with the assistance of spin-orbit coupling effect, gives rise to a nontrivial Berry phase. This results in a competition between weak localization (WL) and weak antilocalization (WAL) in the quantum transport of Dirac-like fermions at the KTO (111) interfaces. Notably, this competitive behavior can be effectively tuned by optical gating via a photoexcitation-induced shift of the Fermi level. Our findings demonstrate a controllable platform based on spin-polarized oxide 2DEGs for quantum transport, opening new avenues for spin-orbitronic and topological electronic applications.

cond-mat.str-el↗

arXiv2025

Quantization property of n-Laplacian mean field equation and sharp Moser-Onofri inequality

In this paper, we are concerned with the following $n$-Laplacian mean field equation \[ \left\{ {\begin{array}{*{20}{c}} { - Δ_n u = λe^u} & {\rm in} \ \ Ω, \\ {\ \ \ \ u = 0} &\ {\rm on}\ \partial Ω, \end{array}} \right. \] \[\] where $Ω$ is a smooth bounded domain of $\mathbb{R}^n \ (n\geq 2)$ and $- Δ_n u =-{\rm div}(|\nabla u|^{n-2}\nabla u)$. We first establish the quantization property of solutions to the above $n$-Laplacian mean field equation. As an application, combining the Pohozaev identity and the capacity estimate, we obtain the sharp constant $C(n)$ of the Moser-Onofri inequality in the $n$-dimensional unit ball $B^n:=B^n(0,1)$, $$\mathop {\inf }\limits_{u \in W_0^{1,n}(B^n)}\frac{1}{ n C_n}\int_{B^n} | \nabla u|^n dx- \ln \int_{B^n} {e^u} dx\geq C(n),$$ which extends the result of Caglioti-Lions-Marchioro-Pulvirenti in \cite{Caglioti} to the case of $n$-dimensional ball. Here $C_n=(\frac{n^2}{n-1})^{n-1} ω_{n-1}$ and $ω_{n-1}$ is the surface measure of $B^n$. For the Moser-Onofri inequality in a general bounded domain of $\mathbb{R}^n$, we apply the technique of $n$-harmonic transplantation to give the optimal concentration level of the Moser-Onofri inequality and obtain the criterion for the existence and non-existence of extremals for the Moser-Onofri inequality.

math.AP↗

arXiv2025

Parrot: A Training Pipeline Enhances Both Program CoT and Natural Language CoT for Reasoning

Natural language chain-of-thought (N-CoT) and Program chain-of-thought (P-CoT) have emerged as two primary paradigms for large language models (LLMs) to solve mathematical reasoning problems. Current research typically endeavors to achieve unidirectional enhancement: P-CoT enhanced N-CoT or N-CoT enhanced P-CoT. In this paper, we seek to fully unleash the two paradigms' strengths for mutual enhancement and ultimately achieve simultaneous improvements. We conduct a detailed analysis of the error types across two paradigms, based on which we propose Parrot, a novel training pipeline for mathematical problems: 1) Three target-designed subtasks integrate sequential P-CoT and N-CoT generation. 2) A subtask hybrid training strategy to facilitate natural language semantic transferability. 3) The converted N-CoT auxiliary reward is designed to alleviate the sparse rewards in P-CoT optimization. Extensive experiments demonstrate that Parrot significantly enhances both the performance of N-CoT and P-CoT, especially on N-CoT. Using Parrot SFT, the N-CoT performance of LLaMA2 and CodeLLaMA achieve gains of +21.87 and +21.48 on MathQA over the RL baseline, which is resource-intensive.

cs.CL↗

arXiv2025

Causal Spatio-Temporal Prediction: An Effective and Efficient Multi-Modal Approach

Spatio-temporal prediction plays a crucial role in intelligent transportation, weather forecasting, and urban planning. While integrating multi-modal data has shown potential for enhancing prediction accuracy, key challenges persist: (i) inadequate fusion of multi-modal information, (ii) confounding factors that obscure causal relations, and (iii) high computational complexity of prediction models. To address these challenges, we propose E^2-CSTP, an Effective and Efficient Causal multi-modal Spatio-Temporal Prediction framework. E^2-CSTP leverages cross-modal attention and gating mechanisms to effectively integrate multi-modal data. Building on this, we design a dual-branch causal inference approach: the primary branch focuses on spatio-temporal prediction, while the auxiliary branch mitigates bias by modeling additional modalities and applying causal interventions to uncover true causal dependencies. To improve model efficiency, we integrate GCN with the Mamba architecture for accelerated spatio-temporal encoding. Extensive experiments on 4 real-world datasets show that E^2-CSTP significantly outperforms 9 state-of-the-art methods, achieving up to 9.66% improvements in accuracy as well as 17.37%-56.11% reductions in computational overhead.

cs.LG↗

arXiv2025

MS-BART: Unified Modeling of Mass Spectra and Molecules for Structure Elucidation

Mass spectrometry (MS) plays a critical role in molecular identification, significantly advancing scientific discovery. However, structure elucidation from MS data remains challenging due to the scarcity of annotated spectra. While large-scale pretraining has proven effective in addressing data scarcity in other domains, applying this paradigm to mass spectrometry is hindered by the complexity and heterogeneity of raw spectral signals. To address this, we propose MS-BART, a unified modeling framework that maps mass spectra and molecular structures into a shared token vocabulary, enabling cross-modal learning through large-scale pretraining on reliably computed fingerprint-molecule datasets. Multi-task pretraining objectives further enhance MS-BART's generalization by jointly optimizing denoising and translation task. The pretrained model is subsequently transferred to experimental spectra through finetuning on fingerprint predictions generated with MIST, a pre-trained spectral inference model, thereby enhancing robustness to real-world spectral variability. While finetuning alleviates the distributional difference, MS-BART still suffers molecular hallucination and requires further alignment. We therefore introduce a chemical feedback mechanism that guides the model toward generating molecules closer to the reference structure. Extensive evaluations demonstrate that MS-BART achieves SOTA performance across 5/12 key metrics on MassSpecGym and NPLIB1 and is faster by one order of magnitude than competing diffusion-based methods, while comprehensive ablation studies systematically validate the model's effectiveness and robustness.

cs.LG↗

arXiv2025

Thermal Hall conductivity of electron-doped cuprates: Electrons and phonons

It has recently become clear that phonons generate a sizable thermal Hall effect in cuprates, whether they are undoped, electron-doped or hole-doped (inside the pseudogap phase). At higher doping, where cuprates are reasonably good metals, mobile electrons also generate a thermal Hall effect, the thermal equivalent of the standard electrical Hall effect. Here we show that in the cleanest crystals of the electron-doped cuprate Nd$_{2-x}$Ce$_{x}$CuO$_{4}$, at high doping, the phonon and electron contributions to the thermal Hall conductivity $κ_{\rm {xy}}$ are of comparable magnitude, but of opposite sign. In samples of lower quality, phonons dominate $κ_{\rm {xy}}$, resulting in a negative $κ_{\rm {xy}}$ at all temperatures. The fact that the negative phononic $κ_{\rm {xy}}$ in the metallic state is similar in magnitude and temperature dependence to that found in the insulating state at lower doping rules out any mechanism based on skew scattering of phonons off charged impurities, since a local charge should be screened in the metallic regime. The phononic $κ_{\rm {xy}}$ is found to persist over the entire doping range where antiferromagnetic correlations are known to be significant, suggesting that such correlations may play a role in generating the phonon thermal Hall effect in electron-doped cuprates. If the same mechanism is also at play in hole-doped cuprates, the presence of a phononic $κ_{\rm {xy}}$ below (and only below) the critical doping $p^{\star}$ would be evidence that spin correlations are a property of the pseudogap phase.

cond-mat.str-el↗

arXiv2025

MULTI: Multimodal Understanding Leaderboard with Text and Images

The rapid development of multimodal large language models (MLLMs) raises the question of how they compare to human performance. While existing datasets often feature synthetic or overly simplistic tasks, some models have already surpassed human expert baselines. In this paper, we present MULTI, a Chinese multimodal dataset derived from authentic examination questions. Comprising over 18,000 carefully selected and refined questions, MULTI evaluates models using real-world examination standards, encompassing image-text comprehension, complex reasoning, and knowledge recall. Additionally, We also introduce MULTI-Elite, a 500-question selected hard subset, and MULTI-Extend with more than 4,500 external knowledge context pieces for testing in-context learning capabilities. Our evaluation highlights substantial room for MLLM advancement, with Qwen2-VL-72B achieving a 76.9% accuracy on MULTI and 53.1% on MULTI-Elite leading 25 evaluated models, compared to human expert baselines of 86.1% and 73.1%. MULTI serves not only as a robust evaluation platform but also paves the way for the development of expert-level AI.

cs.CL↗

arXiv2025

TranSUN: A Preemptive Paradigm to Eradicate Retransformation Bias Intrinsically from Regression Models in Recommender Systems

Regression models are crucial in recommender systems. However, retransformation bias problem has been conspicuously neglected within the community. While many works in other fields have devised effective bias correction methods, all of them are post-hoc cures externally to the model, facing practical challenges when applied to real-world recommender systems. Hence, we propose a preemptive paradigm to eradicate the bias intrinsically from the models via minor model refinement. Specifically, a novel TranSUN method is proposed with a joint bias learning manner to offer theoretically guaranteed unbiasedness under empirical superior convergence. It is further generalized into a novel generic regression model family, termed Generalized TranSUN (GTS), which not only offers more theoretical insights but also serves as a generic framework for flexibly developing various bias-free models. Comprehensive experimental results demonstrate the superiority of our methods across data from various domains, which have been successfully deployed in two real-world industrial recommendation scenarios, i.e. product and short video recommendation scenarios in Guess What You Like business domain in the homepage of Taobao App (a leading e-commerce platform with DAU > 300M), to serve the major online traffic.

cs.IR↗