arXiv subjects

Yu Lei

Publications and source records attributed to Yu Lei.

At least 19 recordsLinked to original sources

arXiv2026

Lower central dark matter densities in nearby galaxies than predicted by simulations

Baryonic feedback in hydrodynamical simulations is typically invoked to alleviate the core--cusp problem in dwarf galaxies. Yet baryonic processes also induce adiabatic contraction of dark matter, producing overly steep density profiles and excessively high dark matter fractions in the inner regions of massive galaxies. The dark matter distribution of galaxies across a wide stellar-mass range is therefore a critical test for such simulations, but a comprehensive benchmark has remained absent. Here, we consistently measure the dark matter distribution from galaxy centres out to radii of 20--50 kpc for 136 nearby galaxies that together span the local mass--size relation over the stellar mass interval $10^9$--$10^{11.5}\,M_{\odot}$. We identify central regions with lower dark matter densities relative to $Λ$CDM simulation expectations---whose extent grows from about 10 kpc to $>50$ kpc as stellar mass increases from $10^{10} M_{\odot}$ to $10^{11.5} M_{\odot}$. Although their physical origin remains unclear, these low--dark matter regions are clearly indicated by the data. Our results provide an important observational benchmark for future hydrodynamical simulations that explore alternative dark matter models and feedback processes.

astro-ph.GA↗

arXiv2026

Horizon-specific Expert Fusion for Photovoltaic Power Forecasting

Short-term photovoltaic power forecasting requires models to represent regular solar cycles and weather-driven fluctuations whose importance changes with the forecast horizon. This study develops a hierarchical ensemble that combines temporal neural models, historical analogs, state climatology, and gradient-boosted trees. Solar geometry and numerical weather forecasts describe the expected generation conditions, while horizon-specific convex weights combine complementary predictions. A separate calibration step uses available historical forecast errors to account for recent bias. The framework is evaluated on public PVDAQ data at 15--240-minute horizons and on three GEFCom2014 solar zones at hourly horizons up to four hours. On PVDAQ, the ensemble achieves a daylight capacity-normalized mean absolute error of 4.315%, reducing error by 4.11% relative to full-feature LightGBM and by 6.03% relative to fine-tuned Chronos-2 under identical calibration. Expert-removal experiments identify redundancy within the ensemble. Across three training seeds on GEFCom2014, learned fusion improves upon equal weighting but performs comparably to LightGBM. The results support horizon-specific combination as a useful forecasting strategy while showing that its advantage over strong individual models depends on the dataset and evaluation period.

cs.AI↗

arXiv2026

Ensemble Complexity in Photovoltaic Forecasting

An ensemble can improve photovoltaic forecasts while adding components that contribute little or increase computation. We assess these effects through matched comparisons and ablations of a fixed heterogeneous predictor bank. Hourly experiments use GEFCom2014 and three additional public datasets, with chronological partitions and three seeds. Under retrospective ERA5 assistance, static fusion reduces scaled mean absolute error against matched boosting by 1.11%, 4.41%, and 1.63% on PVDAQ, OPSD, and Ausgrid; only OPSD remains supported after multiple-comparison correction. Weather gating offers no consistent incremental benefit. Exploratory member removals show group-level dependence alongside individual redundancy. A separate, previously inspected fifteen-minute case replaces one neural member with a tree predictor: normalized error falls by 1.72%, but measured inference is slower. These findings support component-wise evaluation with explicit limits on weather availability and test-set reuse.

cs.LG↗

arXiv2026

The exterior Dirichlet problem for special Lagrangian equations

We establish the existence and uniqueness theorem for the exterior Dirichlet problem for the special Lagrangian equation with prescribed asymptotic behavior at infinity, in both the viscosity setting for all the phases and classical setting for the critical and supercritical phases. These results generalize previous work by the second author by removing restrictive assumptions on the asymptotic matrix and improving the decay rate to the order $2-n$. We also solve the interior Dirichlet problem for the critical special Lagrangian equation and, as applications, all the above-mentioned corresponding problems for the three dimensional quadratic Hessian equation without any admissibility condition.

math.AP↗

arXiv2026

The Hölder regularity of div-curl system with anisotropic coefficients

This research examines the regularity of weak solutions to the Div-Curl system with low regularity anisotropic coefficients. The Hölder regularity of the Div-Curl system with one anisotropic coefficient was an unresolved problem raised by Yin in 2016. We have addressed the open problem, and the findings extend to the scenario involving two anisotropic coefficients. We establish the Hölder regularity of the solution when the coefficients is Hölder continuous. Moreover, the degree of Hölder regularity of the solution can be improved if the coefficient has a greater degree of Hölder regularity.

math.AP↗

arXiv2026

Exterior Dirichlet Problems for Hessian Quotient Equations of Mixed Type

We study the exterior Dirichlet problem for the mixed Hessian quotient equation \[ \frac{σ_k(η(D^2 u))}{σ_l(η(D^2 u))} = 1, \] where $η(M) = (\operatorname{tr} M)I - M$. We establish existence and uniqueness of smooth admissible solutions with prescribed quadratic asymptotics at infinity, and obtain full derivative decay of the remainder. The proof relies on a three-stage subsolution construction.

math.AP↗

arXiv2026

The Illusion of Safety: Multi-Tier Verification of AI vs. Human C++ Code

As large language models (LLMs) are increasingly deployed for systems programming, their ability to generate secure C++ code, where a single memory-safety failure creates an exploitable vulnerability, remains a critical concern. Yet most security evaluations of AI-generated code rely on static analysis alone, which flags warnings without confirming run- time violations or reasoning about untested paths. This study investigates whether AI-generated C++ is measurably less safe than human-written code, and whether common verification tools agree on the risk. We introduce VULBENCH-CPP, a benchmark of 8,918 C++ programs from three open-weight LLMs (Gemma 3 27B IT, LLaMA 3.3 70B Instruct, Qwen 2.5 Coder 32B Instruct) and human authors across 851 competitive-programming tasks. Each program is annotated by four verification tiers: functional testing, static analysis (cppcheck, clang-tidy), dynamic analysis (ASan/UBSan), and bounded model checking (ESBMC). Account- ing for the correlation among solutions to a shared task, we find that AI-generated code is roughly twice as likely as human code to trigger a confirmed runtime violation, even after controlling for code length and test pass-rate. Under static analysis the two look equally safe, but this is misleading: the apparent similarity reflects code length rather than real safety, and the tiers detect largely different classes of violation, demonstrating that no single tier is sufficient. These vulnerability patterns remain consistent across independent generations. We release the benchmark, harness, and annotated results.

cs.SE↗

arXiv2026

From Context to Skills: Can Language Models Learn from Context Skillfully?

Many real-world tasks require language models (LMs) to reason over complex contexts that exceed their parametric knowledge. This calls for context learning, where LMs directly learn relevant knowledge from the given context. An intuitive solution is inference-time skill augmentation: extracting the rules and procedures from context into natural-language skills. However, constructing such skills for context learning scenarios faces two challenges: the prohibitive cost of manual skill annotation for long, technically dense contexts, and the lack of external feedback for automated skill construction. In this paper, we propose Ctx2Skill, a self-evolving framework that autonomously discovers, refines, and selects context-specific skills without human supervision or external feedback. At its core, a multi-agent self-play loop has a Challenger that generates probing tasks and rubrics, a Reasoner that attempts to solve them guided by an evolving skill set, and a neutral Judge that provides binary feedback. Crucially, both the Challenger and the Reasoner evolve through accumulated skills: dedicated Proposer and Generator agents analyze failure cases and synthesize them into targeted skill updates for both sides, enabling automated skill discovery and refinement. To prevent adversarial collapse caused by increasingly extreme task generation and over-specialized skill accumulation, we further introduce a Cross-time Replay mechanism that identifies the skill set achieving the best balance across representative cases for the Reasoner side, ensuring robust and generalizable skill evolution. The resulting skills can be plugged into any language model to obtain better context learning capability. Evaluated on four context learning tasks from CL-bench, Ctx2Skill consistently improves solving rates across backbone models.

cs.AI↗

arXiv2026

Structural properties of one-dimensional $\mathrm{Cs}_2\mathrm{CoCl}_4$ confined within single-walled carbon nanotubes

Crystals under one-dimensional (1D) confinement are well-known to exhibit drastic changes in metallicity, magnetic properties and chemical state, however, the intermediate phase space between binary metal halides and ternary metal halide perovskites remains poorly explored, especially in the context of the rich polymorphism exhibited by both families in the one-dimensional limit. Through aberration-corrected (scanning) transmission electron microscopy and multislice simulations, it is shown that the metal halide $\mathrm{Cs}_2\mathrm{CoCl}_4$ crystallizes in the tetragonal $\wp4/mcc$ and orthorhombic $\wp{mcm}$ rod groups under radial compression within single-walled carbon nanotubes (SWCNTs) of increasingly small diameter, with a massive re-entrant orthorhombic strain towards the $1$ $\mathrm{nm}$ extremum. The persistence of $\mathrm{Co}^{2+}$ is determined from fits to the d.c. magnetization, with a surprisingly small increase in the effective moment ($4.607(3)$ to $4.788(3) \mathit{μ}_\mathrm{B}/\mathrm{f.u.}$) and Weiss constant ($-7.9(3)$ to $-4.09(7) \mathrm{K}$) after confinement in the SWCNTs, suggesting that the confined structure topologically preserves the core magnetic properties of the bulk. Both unconventional polymorphs observed are noticeably different to the high-pressure piezochromic polymorph previously shown to undergo a tetrahedral-to-octahedral coordination transition, highlighting 1D confinement as a unique tool for structural manipulation.

cond-mat.mtrl-sci↗

arXiv2026

Uncovering the dark matter distribution by combining stellar kinematics and integrated HI spectra: Method validation

We determined the dark matter (DM) distribution in galaxies by jointly modelling stellar kinematics from integral field unit (IFU) observations and the gaseous kinematics encoded in a single integrated HI spectrum. The stellar kinematics are described by a triaxial orbit-superposition Schwarzschild model, while the HI gas is described by an idealised disc model; both are governed by the same gravitational potential. The potential comprises the stellar mass, a generalised Navarro-Frenk-White (gNFW) DM halo, and a central black hole. We validated the method on 58 simulated galaxies generated from the TNG50 cosmological simulation. For each galaxy, we created two versions of mock data with azimuthal angles viewed side-on and end-on, thus yielding 116 mock observations in total. Our model recovers the total mass, stellar mass, and DM mass profiles within the data range; the median DM mass of the 58 simulated galaxies is recovered with a relative systematic bias smaller than 20% across all radii from 2--20 kpc. The statistical uncertainties on the DM masses within 5 kpc remain similar to those found with the model constrained by IFU data only. In contrast, the relative uncertainty on the DM mass in the outer regions decreases when the HI spectrum is included; at 20 kpc, it drops markedly, from about 85% to roughly 30%. The DM density slope defined explicitly in the gNFW model is systematically underestimated and thus does not yield a reliable quantity from observations using our approach. Instead, we introduce density slopes evaluated between 2 and 20 kpc, which are statistically well recovered for both the total mass and the DM mass. We demonstrate the reliability of this method in uncovering the DM distribution and emphasise its promise for application to large samples of observed galaxies.

astro-ph.GA↗

arXiv2026

TabKD: Tabular Knowledge Distillation through Interaction Diversity of Learned Feature Bins

Data-free knowledge distillation enables model compression without original training data, critical for privacy-sensitive tabular domains. However, existing methods does not perform well on tabular data because they do not explicitly address feature interactions, the fundamental way tabular models encode predictive knowledge. We identify interaction diversity, systematic coverage of feature combinations, as an essential requirement for effective tabular distillation. To operationalize this insight, we propose TabKD, which learns adaptive feature bins aligned with teacher decision boundaries, then generates synthetic queries that maximize pairwise interaction coverage. Across 4 benchmark datasets and 4 teacher architectures, TabKD achieves highest student-teacher agreement in 14 out of 16 configurations, outperforming 5 state-of-the-art baselines. We further show that interaction coverage strongly correlates with distillation quality, validating our core hypothesis. Our work establishes interaction-focused exploration as a principled framework for tabular model extraction.

cs.LG↗

arXiv2026

TabChange: Precise Attribute Changes in Tabular Data

Modifying an attribute in tabular data often introduces an unnatural instance by breaking its relationships with other attributes. The modified instance must be both natural and minimally changed from the original instance. This paper addresses the challenge of generating such a modified instance. We identify key limitations in existing approaches: generative models either don't support instance-level attribute editing or, in the case of methods like CVAE, retain attribute information in the latent space, leading to unnecessary modifications. To solve this, we propose TabChange, an approach that analyzes the relationship between the attribute of interest and other attributes in the dataset. If the relationship is weak, it simply flips the attribute; if it is strong, it uses an adversarial framework that removes information about the attribute in the latent space representation. This removal enables precise modifications, making only the necessary adjustments to maintain naturalness. Our experiments across seven datasets show that TabChange generates counterfactuals in attributes that are comparable in naturalness and are more proximal to their original instances. This leads to a higher number of valid counterfactuals and a lower number of invalid counterfactuals compared to the baselines.

cs.LG↗

arXiv2026

Are LLMs Socially Adaptive? Contrasting Belief Evolution in Large Language Models and Humans

As large language models (LLMs) increasingly engage in complex social interactions, ensuring that their behaviors align with human ethical principles and intentions, known as value alignment, has become a critical scientific challenge. Existing benchmarks often rely on static assessments and fail to capture the longitudinal dynamics of decision-making or the latent cognitive processes driving agent behavior. In this work, we propose FairMindSim, a realistic simulation benchmark rooted in social psychology that evaluates alignment through continuous economic games. To move beyond black-box observations, we introduce the Belief-Reward Alignment Behavior Evolution Model (BREM), a probabilistic framework that formalizes decision-making as a dynamic trade-off between maximizing extrinsic rewards and upholding intrinsic beliefs. We conducted a large-scale comparative study involving 1,017 human participants and ten LLMs, including GPT-5 and Gemini-3-Pro. Our experimental results reveal a capability linked non linear empirical trend in the Third Party Punishment (TPP) game. Mid capability models exhibit rigid and algorithmic aggression that is characterized by over punishment, while frontier models show a convergence of restraint and a shift toward human like leniency as reasoning capabilities scale. Furthermore, using BREM, we decompose agents longitudinal decision dynamics and find that more advanced models better balance conflicting objectives by reducing belief action inconsistency. Our contributions provide a standardized protocol for psychological stress testing and an interpretable mechanism for analyzing the longitudinal evolution of AI alignment in controlled social dilemma settings.

cs.CE↗

arXiv2026

Threshold-Guided Optimization for Visual Generative Models

Aligning large visual generative models with human feedback is often performed through pairwise preference optimization. While such approaches are conceptually simple, they fundamentally rely on annotated pairs, limiting scalability in settings where feedback is collected as independent scalar ratings. In this work, we revisit the KL-regularized alignment objective and show that the optimal policy implicitly compares each sample's reward to an instance-specific baseline that is generally intractable. We propose a threshold-guided alignment framework that replaces this oracle baseline with a data-driven global threshold estimated from empirical score statistics. This formulation turns alignment into a binary decision task on unpaired data, enabling effective optimization directly from scalar feedback. We also incorporate a confidence weighting term to emphasize samples whose scores deviate strongly from the threshold, improving sample efficiency. Experiments across both diffusion and masked generative paradigms, spanning three test sets and five reward models, show that our method consistently improves preference alignment over previous methods. These results position our threshold-guided framework as a simple yet principled alternative for aligning visual generative models without paired comparisons.

cs.LG↗

arXiv2026

ReaGeo: Reasoning-Enhanced End-to-End Geocoding with LLMs

This paper proposes ReaGeo, an end-to-end geocoding framework based on large language models, designed to overcome the limitations of traditional multi-stage approaches that rely on text or vector similarity retrieval over geographic databases, including workflow complexity, error propagation, and heavy dependence on structured geographic knowledge bases. The method converts geographic coordinates into geohash sequences, reformulating the coordinate prediction task as a text generation problem, and introduces a Chain-of-Thought mechanism to enhance the model's reasoning over spatial relationships. Furthermore, reinforcement learning with a distance-deviation-based reward is applied to optimize the generation accuracy. Comprehensive experiments show that ReaGeo can accurately handle explicit address queries in single-point predictions and effectively resolve vague relative location queries. In addition, the model demonstrates strong predictive capability for non-point geometric regions, highlighting its versatility and generalization ability in geocoding tasks.

cs.AI↗

arXiv2026

A novel LSTM music generator based on the fractional time-frequency feature extraction

In this paper, we propose a novel approach for generating music based on an artificial intelligence (AI) system. We analyze the features of music and use them to fit and predict the music. The fractional Fourier transform (FrFT) and the long short-term memory (LSTM) network are the foundations of our method. The FrFT method is used to extract the spectral features of a music piece, where the music signal is expressed on the time and frequency domains. The LSTM network is used to generate new music based on the extracted features, where we predict the music according to the hidden layer features and real-time inputs using GiantMIDI-Piano dataset. The results of our experiments show that our proposed system is capable of generating high-quality music that is comparable to human-generated music.

cs.SD↗

arXiv2026

A Unified Framework for Modeling Heterogeneous Financial Data via Dual-Granularity Prompting

Recent industrial credit scoring models remain heavily reliant on manually tuned statistical learning methods. Despite their potential, deep learning architectures have struggled to consistently outperform traditional statistical models in industrial credit scoring, largely due to the complexity of heterogeneous financial data and the challenge of modeling evolving creditworthiness. To bridge this gap, we introduce FinLangNet, a novel framework that reformulates credit scoring as a multi-scale sequential learning problem. FinLangNet processes heterogeneous financial data through a dual-module architecture that combines tabular feature extraction with temporal sequence modeling, generating probability distributions of users' future financial behaviors across multiple time horizons. A key innovation is our dual-prompt mechanism within the sequential module, which introduces learnable prompts operating at both feature-level granularity for capturing fine-grained temporal patterns and user-level granularity for aggregating holistic risk profiles. Notably, real world deployment yielded a 6.3 pp improvement in KS, along with a 9.9\% reduction in bad debt rate.

cs.CE↗

arXiv2026

A Mechanistic Analysis of Sim-and-Real Co-Training in Generative Robot Policies

Co-training, which combines limited in-domain real-world data with abundant surrogate data such as simulation or cross-embodiment robot data, is widely used for training generative robot policies. Despite its empirical success, the mechanisms that determine when and why co-training is effective remain poorly understood. We investigate the mechanism of sim-and-real co-training through theoretical analysis and empirical study, and identify two intrinsic effects governing performance. The first, \textbf{``structured representation alignment"}, reflects a balance between cross-domain representation alignment and domain discernibility, and plays a primary role in downstream performance. The second, the \textbf{``importance reweighting effect"}, arises from domain-dependent modulation of action weighting and operates at a secondary level. We validate these effects with controlled experiments on a toy model and extensive sim-and-sim and sim-and-real robot manipulation experiments. Our analysis offers a unified interpretation of recent co-training techniques and motivates a simple method that consistently improves upon prior approaches. More broadly, our aim is to examine the inner workings of co-training and to facilitate research in this direction.

cs.RO↗