arXiv subjects

Fei Wu

Publications and source records attributed to Fei Wu.

At least 19 recordsLinked to original sources

arXiv2026

Efficient Covariance-Aware Vicinal Latent Diffusion for Continuous Conditional Image Generation

Continuous Conditional Generative Modeling (CCGM) aims to synthesize high-dimensional data conditioned on continuously valued scalar labels. However, existing continuous diffusion models face severe data sparsity under imbalanced and long-tailed label distributions, which can lead to visual artifacts and condition inconsistency. Moreover, inference-time classifier-free guidance and dense pixel-space integration introduce substantial sampling overhead, limiting practical deployment. To address these challenges, we propose CAViLD, an efficient latent diffusion framework for continuous conditional generation. CAViLD first constructs a purified condition-aware latent space through density-adaptive vicinal learning, auxiliary semantic regularization, and NIQE-guided negative data augmentation. It then internalizes the conditional guidance direction through an EMA teacher, enabling single-branch conditional inference, and introduces covariance-aware progressive distillation for label-dependent probability-flow ODEs, supporting sampling down to four NFEs. Experiments on RC-49, UTKFace, and Steering Angle demonstrate a favorable trade-off among generation fidelity, continuous condition consistency, and sampling efficiency.

cs.CV↗

arXiv2026

More Meters or More Prices? One- vs. Multi-dimensional Fee Mechanisms on Ethereum

Permissionless blockchains use transaction fee mechanisms to allocate the resources that transactions consume, such as execution, data bandwidth, and state creation. Recent blockchain designs explore pricing these resources separately rather than through a single unit. We study this shift on Ethereum, whose roadmap moves in stages from a one-dimensional fee mechanism to a multi-dimensional one. We build a demand model for the three resources, calibrated from Ethereum's 2025 block-capacity increases, and find that state creation is the most price-responsive. We then evaluate both mechanisms at higher block capacity, combining the model with transaction-level metering reconstruction and block-level simulation. Under a one-dimensional mechanism, state creation determines the fee and suppresses execution throughput. Adjusting how state is metered can align resource usage under constant demand, but a single fee cannot track all resources when demand varies, leaving all underutilized. A multi-dimensional mechanism removes this fee coupling, but execution still generates data that must be propagated. Across different allocations between block execution and propagation time, the multi-dimensional mechanism delivers 4--34\% more execution throughput at similarly low congestion, with higher target utilization and state growth near its intended level.

cs.GT↗

arXiv2026

ICIBS: Increasing the Context for Ice Boundary Segmentation

Mountain glaciers and ice sheets are undergoing rapid loss of ice mass. To assess their current volume and ice-thickness distribution, and to accurately forecast their future, ground-penetrating radar systems are employed to measure ice thickness. These instruments produce radio-echo sounding data that reveal the internal structure of the ice and its boundaries at the surface and beneath the ice body. However, delineating these interfaces is time-consuming as the surveyed area can span several hundred kilometers. To accelerate the labelling process, automated methods have emerged, but they often struggle to delineate the ice bottom, particularly in temperate glaciers. The high water content inherent in temperate ice increases both scattering and attenuation of the radar signal, resulting in weak and ambiguous reflections at the ice bottom and generally noisier data. As a result, the visual features corresponding to the ice bottom become ambiguous. To address this issue, we propose several extensions to deep learning models that incorporate additional contextual information into the prediction. Specifically, we integrate modifications that incorporate additional spatial and geo-spatial contextual information into the network. With these enhancements, we achieve a new state-of-the-art performance on the IceAnatomy benchmark dataset, with mean errors of 10.9 m for the FAU subset and 43.4 m for the CReSIS subset in a single combined system. We also demonstrate that our method generalises well to previously unseen data, outperforming the baseline by 49.6 % on new campaign data from Patagonia. The model code and the new dataset are available at https://github.com/ki7077/ICIBS/tree/main and https://zenodo.org/records/23392580.

cs.CV↗

arXiv2026

VASPO: Self-Distillation with Privileged Value Estimation

On-policy self-distillation provides dense supervision for reasoning models via a privileged self-teacher. However, this supervision reflects local token preferences rather than the expected return of an unguided student continuation. Recent methods combine this signal with GRPO. They use GRPO for update directions and the teacher for signal scaling. However, we demonstrate a more effective integration within PPO. Specifically, we identify complementary positional dynamics: privileged information enhances a critic's value-prediction accuracy toward the response's end, whereas self-teacher guidance is strong initially, drops mid-generation, and rebounds finally. These insights motivate Value-Augmented Self-Distillation Policy Optimization (VASPO), wherein a self-teacher and critic evaluate identical privileged information at student-generated prefixes. The critic computes token-level PPO advantages from predicted continuation returns, while the self-teacher modulates selected positive advantages via bounded probability ratios. Across six mathematical reasoning benchmarks, VASPO improves mean accuracy over PPO by 2.59 and 1.81 percentage points on Qwen3-4B and OLMo3-7B variants, respectively, and achieves the highest mean accuracy among evaluated methods on Knights and Knaves. Our code is available at https://github.com/cii030/VASPO.

cs.CL↗

arXiv2026

TopoCurate:Modeling Interaction Topology for Tool-Use Agent Training

Training tool-use agents typically relies on outcome-based filtering: Supervised Fine-Tuning (SFT) on successful trajectories and Reinforcement Learning (RL) on pass-rate-selected tasks. However, this paradigm ignores interaction dynamics: successful trajectories may lack error recovery or exhibit redundancy, while pass rates fail to distinguish structurally informative tasks from trivial ones. We propose \textbf{TopoCurate}, an interaction-aware framework that projects multi-trial rollouts from the same task into a unified semantic quotient topology. By merging equivalent action-observation states, this projection transforms scattered linear trajectories into a structured manifold that explicitly captures how tool invocations and environmental responses drive the divergence between effective strategies and failure modes. Leveraging this representation, we introduce a dual-selection mechanism: for SFT, we prioritize trajectories demonstrating reflective recovery, semantic efficiency, and strategic diversity to mitigate covariate shift and mode collapse; for RL, we select tasks with high error branch ratios and strategic heterogeneity, maximizing gradient Signal-to-Noise Ratio to address vanishing signals in sparse-reward settings. Evaluations on BFCLv3 and Tau2 Bench show that TopoCurate achieves consistent gains of 4.2\% (SFT) and 6.9\% (RL) over state-of-the-art baselines. We will release the code and data soon for further investigations.

cs.LG↗

arXiv2026

SOAR: Source-Aware Reinforcement Learning with Precise Credit Assignment for Search Agents

Reinforcement learning (RL) for search agents often suffers from evidence fabrication, a failure mode where agents bypass external information ($E_{ext}$) and fabricate an incorrect answer. We argue that this fabrication arises from an over-reliance on unreliable internal knowledge ($E_{int}$) as shortcuts due to the inherent opacity of reasoning trajectories. To address this, we propose SOAR (SOurce-Aware Reinforcement learning), a paradigm that transforms monolithic reasoning into source-explicit trajectories. SOAR (i) disentangles $E_{ext}$ and $E_{int}$ into parallel, observable streams via a Source-Aware Workflow, and (ii) incorporates Source-Consistent Reward to guide the policy evolution. By making information-source attribution explicit, SOAR enables precise credit assignment between different knowledge streams. Evaluations show that SOAR outperforms baselines on 3B and 7B backbones by 7.3% and 5.4%, while reducing evidence fabrication by 74.8%. These confirm that SOAR significantly improves performance on knowledge-intensive benchmarks while effectively mitigating evidence fabrication, fostering more faithful and grounded reasoning with enhanced robustness.

cs.CL↗

arXiv2026

InfiFPO: Implicit Model Fusion via Preference Optimization in Large Language Models

Model fusion combines multiple Large Language Models (LLMs) with different strengths into a more powerful, integrated model through lightweight training methods. Existing works on model fusion focus primarily on supervised fine-tuning (SFT), leaving preference alignment (PA) --a critical phase for enhancing LLM performance--largely unexplored. The current few fusion methods on PA phase, like WRPO, simplify the process by utilizing only response outputs from source models while discarding their probability information. To address this limitation, we propose InfiFPO, a preference optimization method for implicit model fusion. InfiFPO replaces the reference model in Direct Preference Optimization (DPO) with a fused source model that synthesizes multi-source probabilities at the sequence level, circumventing complex vocabulary alignment challenges in previous works and meanwhile maintaining the probability information. By introducing probability clipping and max-margin fusion strategies, InfiFPO enables the pivot model to align with human preferences while effectively distilling knowledge from source models. Comprehensive experiments on 11 widely-used benchmarks demonstrate that InfiFPO consistently outperforms existing model fusion and preference optimization methods. When using Phi-4 as the pivot model, InfiFPO improve its average performance from 79.95 to 83.33 on 11 benchmarks, significantly improving its capabilities in mathematics, coding, and reasoning tasks.

cs.LG↗

arXiv2026

MetaOPD: Meta-Learned Token Weighting for On-Policy Distillation

On-policy distillation (OPD) trains a student on its own generated responses using token-level teacher supervision. However, uniform weighting overlooks differences in token learning value, while existing weighting methods rely on predefined mappings from prediction signals to token weights. These mappings are not learned from the effectiveness of the resulting student updates, limiting their ability to adapt to evolving learning needs. In this paper, we propose MetaOPD, a bilevel optimization framework that jointly learns the student model and a lightweight token-weighting network. The inner objective updates the student through weighted OPD, while the outer objective optimizes the weighting network using validation loss on reference solutions after a virtual student update. Differentiating through this update connects weighting decisions to their effects on post-update performance, allowing the mapping from prediction signals to token weights to evolve alongside the student. Experiments on six mathematical reasoning and three out-of-domain datasets, covering two student scales and seven baselines, demonstrate the effectiveness of MetaOPD, with Avg@8/Pass@8 gains over OPD of 1.99/5.97 percentage points for the 0.6B student and 2.25/6.41 points for the 1.7B student.

cs.AI↗

arXiv2026

Sign-changing solutions for the critical Neumann problem in the three-dimensional unit ball

We consider the existence of nonradial sign-changing solutions to the problem $Δu-μu+|u|^4u=0$ with zero Neumann boundary conditions in the unit ball for an arbitrary fixed parameter $μ>0$. We prove that, for every $μ>0$, this problem admits infinitely many distinct nonradial sign-changing solutions. More precisely, for every sufficiently large even integer $K$, we construct a solution by gluing $K$ symmetrically arranged, suitably transformed copies of a sign-changing entire solution of the critical equation in $\mathbb{R}^3$. The construction combines an inner--outer gluing scheme with a finite-dimensional reduction. A detailed expansion of the reduced energy allows us to locate an interior critical point in the admissible parameter region and thereby complete the construction.

math.AP↗

arXiv2026

Standardized Dynamic Borrowing: Uniform Control of Information Borrowing

Dynamic borrowing can improve efficiency in clinical trials using external information, but existing methods may not offer transparent, prespecified control over borrowing. We introduce Standardized Dynamic Borrowing (SDB), which combines a source-informed component with a no-borrowing component and updates their relative odds using a Bayes factor standardized by its supremum over admissible target data. The posterior density under the no-borrowing component is obtained by normalizing the target-data likelihood. The prespecified source-component initial weight is a sharp uniform upper bound on its posterior weight, providing an interpretable borrowing constraint. When source and target models belong to the same one-parameter canonical exponential family, posterior expected local-information-ratio effective sample size equals target sample size plus borrowed source-prior effective sample size minus a nonnegative heterogeneity penalty. For normally distributed outcomes, borrowing depends on a standardized source-target conflict $Z$-score and decreases according to a standard normal kernel. We also jointly select the initial weight and target sample size to meet prespecified power and Type I error requirements. Normal and binomial illustrations demonstrate how SDB translates source-target compatibility into data-adaptive borrowing and how the calibration procedure prespecifies the source-component initial weight and target sample size under operating-characteristic constraints. These results provide a transparent framework for prespecifying, controlling, and interpreting dynamic borrowing.

stat.ME↗

arXiv2026

BrainWave: A Brain Signal Foundation Model for Clinical Applications

Neural electrical activity is fundamental to brain function, underlying a range of cognitive and behavioral processes, including movement, perception, decision-making, and consciousness. Abnormal patterns of neural signaling often indicate the presence of underlying brain diseases. The variability among individuals, the diverse array of clinical symptoms from various brain disorders, and the limited availability of diagnostic classifications, have posed significant barriers to formulating reliable model of neural signals for diverse application contexts. Here, we present BrainWave, the first foundation model for both invasive and non-invasive neural recordings, pretrained on more than 40,000 hours of electrical brain recordings (13.79 TB of data) from approximately 16,000 individuals. Our analysis show that BrainWave outperforms all other competing models and consistently achieves state-of-the-art performance in the diagnosis and identification of neurological disorders. We also demonstrate robust capabilities of BrainWave in enabling zero-shot transfer learning across varying recording conditions and brain diseases, as well as few-shot classification without fine-tuning, suggesting that BrainWave learns highly generalizable representations of neural signals. We hence believe that open-sourcing BrainWave will facilitate a wide range of clinical applications in medicine, paving the way for AI-driven approaches to investigate brain disorders and advance neuroscience research.

q-bio.NC↗

arXiv2026

PulseInfer: I/O-Centric Sparse KV Cache Offloading for Efficient Long-Context LLM Decoding

Long-context LLM serving is increasingly bottlenecked by decode, where large KV caches limit batch size and underutilize GPUs. Sparse KV cache offloading expands effective capacity by storing most historical KV blocks in CPU DRAM and recalling only selected blocks on demand. However, we find that existing offloading systems shift the bottleneck to CPU-GPU recall I/O: recall volume varies widely across layers, decode steps and requests, while headwise sparse selection fragments recalls into many small PCIe transfers. This paper presents PulseInfer, an I/O-centric sparse KV cache offloading system. PulseInfer hides variable recall latency with interruptible layer-wise scheduling, adapts offloading decisions with IO-Adaptive Offloading Admission, and coalesces fragmented transfers using SoloHead sparse selection and a gather-scatter I/O engine. Implemented on SGLang, PulseInfer improves decode throughput by up to 4.7x over SGLang and 2.6x over the best existing offloading baseline, while reducing TPOT by up to 76% and preserving near-lossless accuracy.

cs.LG↗

arXiv2026

Preventing Rank Collapse in Federated Low-Rank Adaptation with Client Heterogeneity

Federated low-rank adaptation (FedLoRA) has facilitated communication-efficient and privacy-preserving fine-tuning of foundation models for downstream tasks. In practical federated learning scenarios, client heterogeneity in system resources and data distributions motivates the use of heterogeneous LoRA ranks across clients. However, we identify a previously overlooked phenomenon in heterogeneous FedLoRA with SVD-based allocation, termed rank collapse, where the energy of the global update becomes concentrated in the minimum shared rank, resulting in suboptimal performance and high sensitivity to rank configurations. Through theoretical analysis, we reveal the underlying mechanism of rank collapse: a mismatch between rank-agnostic aggregation weights and rank-dependent client contributions, which systematically suppresses higher-rank updates at a geometric rate over rounds. Motivated by this insight, we propose raFLoRA, a rank-partitioned aggregation method that decomposes local updates into rank partitions and then aggregates each partition weighted by its effective client contributions. Extensive experiments across vision, language, and reasoning tasks show that raFLoRA prevents rank collapse, improves model performance, and enhances robustness across diverse heterogeneous configurations compared with strong FedLoRA baselines.

cs.LG↗

arXiv2026

Physics-Informed Neural Networks for Static Black-Hole Exterior Metrics: Charge and Cosmological-Constant Sweeps

We apply physics-informed neural networks (PINNs) to recover the time-time component of static, spherically symmetric black-hole exterior metrics from a reduced ordinary differential equation (ODE). Inspired by recent work on solving Einstein field equations with deep learning~\cite{Li2023}, we encode the vacuum/charged exterior through a residual loss and an asymptotic boundary constraint---not by embedding analytic metric terms such as $2M/r$ directly into the network output. Unlike the distributed PINN (DPINN) strategy of Ref.~\cite{Li2023}, which partitions the radial domain into subdomains with separate networks, we employ a \emph{unified} fully connected network over the entire interval $[r_{\min},r_{\max}]$, avoiding spurious jumps at subdomain interfaces. Moreover, whereas Ref.~\cite{Li2023} restricts training to $r\in(10,300)$---well outside the steep $1/r$ and $1/r^2$ curvature of the inner exterior---we begin at $r_{\min}=10^{-2}M$, spanning three decades in radius and covering the strongly varying region that DPINNs sidestep by domain truncation. Holding the mass fixed at $M=1$, we sweep electric charge $Q\in\{0,0.5,1.0,1.1\}$ at $Λ=0$ and sweep $Λ\in\{0,0.1,-0.1\}$ at $Q=0.5$. All configurations achieve relative $\mathcal{L}_2$ errors below $6\%$ against the analytic reference. For the representative case $Q=0.5$, $Λ=0.1$, three independent trainings with fixed seeds yield relative errors of $3.89\%$, $1.08\%$, and $2.76\%$ (mean $2.58\%$, standard deviation $1.41\%$), demonstrating robustness of the mesh-free approach without labeled field data.

gr-qc↗

arXiv2026

Towards Efficient Reasoning: Learning Causal Shortcuts for Diffusion Language Models

Diffusion Language Models (DLMs) have attracted significant attention for their strong reasoning ability. However, under a bidirectional attention mechanism, DLMs operate over an exponentially large exploration space compared to autoregressive models (ARMs), making it challenging to focus on reasoning-guiding tokens under random masking. We define causal shortcuts as token chains that cover the full sequence and provide explicit guidance towards correct reasoning trajectories. We analyze the effects of causal shortcuts on the reasoning accuracy and convergence speed of DLMs, and find that they largely improve answer convergence efficiency and generation accuracy. Motivated by this, we propose a Causal Shortcut Learning (CSL) Framework for DLMs. Specifically, we introduce a step-by-step token extraction procedure to extract causal shortcuts from data, and apply parallel prioritized masking on these tokens during training to enable efficient and accurate convergence to correct answers via causal shortcuts. Extensive experiments across multiple reasoning benchmarks and two base models demonstrate that CSL consistently outperforms existing SFT-variant baselines, achieving an average improvement of $1.92\%$ over SFT-only models, and up to $4.20\%$ on MATH-500. The code is available at the \href{https://github.com/ZJUDianJin/Causal-Shortcuts-Learning}{https://github.com/ZJUDianJin/Causal-Shortcuts-Learning

cs.CL↗

arXiv2026

Decoupled Causal Discovery

Causal discovery from observational data is a fundamental yet challenging task in scientific research. While existing approaches are primarily based on conditional independence tests, structure scores, or restrictive functional assumptions, we propose Decoupled Causal Discovery (DCD), a novel decoupling-based perspective that does not rely on these methodologies. DCD directly identifies the Markov boundary (MB) by decoupling non-target variables via weighting functions, such that only variables within the MB preserve dependence with the target under the decoupled distribution. Building on this, DCD iteratively constructs the Completed Partially Directed Acyclic Graph (CPDAG) by exploiting structural asymmetries within the MBs. We establish the theoretical identifiability, soundness, and completeness of DCD. Empirical evaluations demonstrate that DCD achieves strong performance, particularly excelling in challenging noise regimes.

cs.LG↗

arXiv2026

TTGBench: Benchmarking Topological Evolution and Semantic Drift in Text-attributed Temporal Graphs

Temporal graph learning models the evolution of dynamic systems, where both structural interactions and semantic states change over time. However, existing benchmarks primarily emphasize structural evolution via temporal link prediction (TLP), while support for semantic evolution remains limited. Although temporal node classification (TNC) is sometimes included, it is typically restricted to simplistic binary settings that fail to capture realistic semantic drift. Moreover, commonly used datasets exhibit high link repetition, leading to inflated performance estimates and obscuring true model capability. To address these limitations, we introduce \textbf{TTGBench}, a new benchmark that jointly evaluates structural and semantic evolution. TTGBench comprises six real-world, text-rich datasets characterized by \emph{Dual Volatility}, enabling rigorous and fair evaluation of existing models. Notably, it is the first benchmark to support both multi-class and multi-label TNC, filling a critical gap in evaluating temporal semantic drift. We conduct a comprehensive evaluation of 17 state-of-the-art methods across Temporal Graph Neural Networks (TGNNs) and Large Language Model (LLM)-based paradigms. The results reveal a clear \emph{capability divide} between the two paradigms: TGNN-based methods excel at structural prediction but fail at semantic tracking, whereas LLM-based predictors show the opposite trend. Through in-depth analysis, we uncover their fundamental limitations and provide insights for developing more comprehensive temporal graph models.

cs.AI↗

arXiv2026

Robust Deterministic Policies for Markov Decision Processes under Budgeted Uncertainty

This paper studies the computation of robust deterministic policies for Markov Decision Processes (MDPs) in the Lightning Does Not Strike Twice (LDST) model of Mannor, Mebel and Xu (ICML '12). In this model, designed to provide robustness in the face of uncertain input data while not being overly conservative, transition probabilities and rewards are uncertain and the uncertainty set is constrained by a budget that limits the number of states whose parameters can deviate from their nominal values. Mannor et al. (ICML '12) showed that optimal randomized policies for MDPs in the LDST regime can be efficiently computed when only the rewards are affected by uncertainty. In contrast to these findings, we observe that the computation of optimal deterministic policies is $N\!P$-hard even when only a single terminal reward may deviate from its nominal value and the MDP consists of $2$ time periods. For this hard special case, we then derive a constant-factor approximation algorithm by combining two relaxations based on the Knapsack Cover and Generalized Assignment problem, respectively. For the general problem with possibly a large number of deviations and a longer time horizon, we derive strong inapproximability results for computing robust deterministic policies as well as $Σ_2^p$-hardness, indicating that the general problem does not even admit a compact mixed integer programming formulation.

math.OC↗