arXiv subjects

Hanwen Zhang

Publications and source records attributed to Hanwen Zhang.

At least 37 records · Page 2Linked to original sources

arXiv2026

Contextualized Code Pretraining for Code Generation

As code generation becomes increasingly central to improving software development efficiency, modern code models are largely trained and evaluated on code with natural-language descriptions. In real projects, developers often implement missing functions under limited project-specific artifacts, while the local call-site context is already available in the surrounding code. This usage context provides actionable cues about expected behavior, but existing models are not explicitly optimized to leverage it reliably, leading to implementations that may not integrate smoothly with surrounding usage in repository settings. In this work, we propose contextualized code pretraining, an invocation-aware framework that integrates calling context into both the training and evaluation of code models. Using static analysis, we automatically extract large-scale caller-callee pairs from real repositories to construct pretraining tasks and benchmarks that condition generation on the calling context. We train CallerGen, the first code models pretrained with invocation-aware objectives spanning multiple sizes, and evaluate them on CallerEval, a new benchmark featuring realistic scenarios. Experiments show that CallerGen outperforms comparable-scale models and remains competitive with larger ones across two benchmarks. Our 220M and 0.5B models achieve 16.58% and 22.81@% pass1, surpassing baselines on CallerEval. These results highlight the importance of calling context in realistic code generation.

cs.SE↗

arXiv2026

Static to Dynamic Correlation Clustering

Correlation clustering is a well-studied problem, first proposed by Bansal, Blum, and Chawla [Mach. Learn. '04]. The input is an unweighted, undirected graph. The problem is to cluster the vertices so as to minimize the number of edges between vertices in different clusters and missing edges between vertices inside the same cluster. This problem has a wide application in data mining and machine learning. We introduce a general framework that transforms existing static correlation clustering algorithms into fully-dynamic ones that work against an adaptive adversary. We show how to apply our framework to known efficient correlation clustering algorithms, starting from the classic 3-approximate Pivot algorithm from Ailon, Charikar and Newman [JACM'08]. Applied to the most recent sublinear $1.485$-approximation algorithm from Cao, Cohen-Addad, Lee, Li, Lolck, Newman, Thorup, Vogl, Yan and Zhang [STOC'25], we get a $1.485$-approximation fully-dynamic algorithm that works with worst-case constant update time. The original static algorithm gets its approximation factor with constant probability, and we get the same against an adaptive adversary in the sense that for any given update step, not known to our algorithm, our solution is a $1.485$-approximation with constant probability when we reach this update. Most of previous dynamic algorithms, including the celebrated result from Behnezhad, Charikar, Ma and Tan [FOCS'19], had approximation factors around $3$ in expectation, and they could only handle an oblivious adversary. A recent algorithm by Braverman, Dharangutte, Pai, Shah, and Wang [AISTATS'25] could handle an adaptive adversary, but it has a large unspecified constant approximation ratio. This contrasts with our general transformation, which works with all the best approximation factors known for the static case.

cs.DS↗

arXiv2026

Decoupling Knowledge and Task Subspaces for Composable Parametric Retrieval Augmented Generation

Parametric Retrieval-Augmented Generation (PRAG) encodes external documents into lightweight parameter modules that can be retrieved and merged at inference time, offering a promising alternative to in-context retrieval augmentation. Despite its potential, many PRAG implementations train document adapters with task-supervised objectives, which may cause each adapter to encode both document-specific facts and reusable task-solving behavior. This entanglement may make adapter composition less reliable: when multiple adapters are merged at inference time, their overlapping task behaviors can accumulate together with document-specific updates, potentially making the merged adapter less stable and less focused on the intended document knowledge. To examine this issue, we explore Orthogonal Subspace Decomposition (OSD), an adapter-training setup that separates reusable task behavior from document-specific knowledge adapters. Concretely, we first train a Task LoRA to capture reusable task behavior, and then train document LoRAs to encode document-specific knowledge in a orthogonal subspace. This setup provides a controlled way to examine how orthogonalizing task and document LoRA updates affects adapter composition in multi-document PRAG. Experiments across multiple knowledge-intensive tasks and model scales suggest that this orthogonalization strategy can improve compositional robustness in parametric RAG, especially when multiple document adapters are merged.

cs.CL↗

arXiv2026

Ouroboros: Single-step Diffusion Models for Cycle-consistent Forward and Inverse Rendering

While multi-step diffusion models have advanced both forward and inverse rendering, existing approaches often treat these problems independently, leading to cycle inconsistency and slow inference speed. In this work, we present Ouroboros, a framework composed of two single-step diffusion models that handle forward and inverse rendering with mutual reinforcement. Our approach extends intrinsic decomposition to both indoor and outdoor scenes and introduces a cycle consistency mechanism that ensures coherence between forward and inverse rendering outputs. Experimental results demonstrate state-of-the-art performance across diverse scenes while achieving substantially faster inference speed compared to other diffusion-based methods. We also demonstrate that Ouroboros can transfer to video decomposition in a training-free manner, reducing temporal inconsistency in video sequences while maintaining high-quality per-frame inverse rendering.

cs.CV↗

arXiv2026

Finding roots of complex analytic functions via generalized colleague matrices

We present a scheme for finding all roots of an analytic function in a square domain in the complex plane. The scheme can be viewed as a generalization of the classical approach to finding roots of a function on the real line, by first approximating it by a polynomial in the Chebyshev basis, followed by diagonalizing the so-called ''colleague matrices''. Our extension of the classical approach is based on several observations that enable the construction of polynomial bases in compact domains that satisfy three-term recurrences and are reasonably well-conditioned. This class of polynomial bases gives rise to ''generalized colleague matrices'', whose eigenvalues are roots of functions expressed in these bases. In this paper, we also introduce a special-purpose QR algorithm for finding the eigenvalues of generalized colleague matrices, which is a straightforward extension of the recently introduced componentwise stable QR algorithm for the classical cases (See [Serkh]). The performance of the schemes is illustrated with several numerical examples.

math.NA↗

arXiv2026

XEmoGPT: An Explainable Multimodal Emotion Recognition Framework with Cue-Level Perception and Reasoning

Explainable Multimodal Emotion Recognition plays a crucial role in applications such as human-computer interaction and social media analytics. However, current approaches struggle with cue-level perception and reasoning due to two main challenges: 1) general-purpose modality encoders are pretrained to capture global structures and general semantics rather than fine-grained emotional cues, resulting in limited sensitivity to emotional signals; and 2) available datasets usually involve a trade-off between annotation quality and scale, which leads to insufficient supervision for emotional cues and ultimately limits cue-level reasoning. Moreover, existing evaluation metrics are inadequate for assessing cue-level reasoning performance. To address these challenges, we propose eXplainable Emotion GPT (XEmoGPT), a novel EMER framework capable of both perceiving and reasoning over emotional cues. It incorporates two specialized modules: the Video Emotional Cue Bridge (VECB) and the Audio Emotional Cue Bridge (AECB), which enhance the video and audio encoders through carefully designed tasks for fine-grained emotional cue perception. To further support cue-level reasoning, we construct a large-scale dataset, EmoCue, designed to teach XEmoGPT how to reason over multimodal emotional cues. In addition, we introduce EmoCue-360, an automated metric that extracts and matches emotional cues using semantic similarity, and release EmoCue-Eval, a benchmark of 400 expert-annotated samples covering diverse emotional scenarios. Experimental results show that XEmoGPT achieves strong performance in both emotional cue perception and reasoning.

cs.MM↗

arXiv2026

Evaluating High-Resolution Piano Sustain Pedal Depth Estimation with Musically Informed Metrics

Evaluation for continuous piano pedal depth estimation tasks remains incomplete when relying only on conventional frame-level metrics, which overlook musically important features such as direction-change boundaries and pedal curve contours. To provide more interpretable and musically meaningful insights, we propose an evaluation framework that augments standard frame-level metrics with an action-level assessment measuring direction and timing using segments of press/hold/release states and a gesture-level analysis that evaluates contour similarity of each press-release cycle. We apply this framework to compare an audio-only baseline with two variants: one incorporating symbolic information from MIDI, and another trained in a binary-valued setting, all within a unified architecture. Results show that the MIDI-informed model significantly outperforms the others at action and gesture levels, despite modest frame-level gains. These findings demonstrate that our framework captures musically relevant improvements indiscernible by traditional metrics, offering a more practical and effective approach to evaluating pedal depth estimation models.

cs.IR↗

arXiv2026

Feasibility-Aware Learning-to-Optimize in Wireless Communication Resource Allocation

The emergence of 6G wireless communication enables massive edge device access and supports real-time intelligent services such as the Internet of things (IoT) and vehicle-to-everything (V2X). However, the surge in edge devices connectivity renders wireless resource allocation (RA) tasks as large-scale constrained optimization problems, whereas the stringent real-time requirement poses significant computational challenge for traditional algorithms. To address the challenge, feasibility-aware learning-to-optimize (L2O) techniques have recently gained attention. These learning-based methods offer efficient alternatives to conventional solvers by directly learning mappings from system parameters to feasible and near-optimal solutions. This article provide a comprehensive review of L2O model designs and feasibility enforcement techniques and investigates the application of constrained L2O in wireless RA systems and. The paper also presents a case study to benchmark different L2O approaches in weighted sum rate problem, and concludes by identifying key challenges and future research directions.

cs.IT↗

arXiv2026

DSFedMed: Dual-Scale Federated Medical Image Segmentation via Mutual Distillation Between Foundation and Lightweight Models

Foundation Models (FMs) have demonstrated strong generalization across diverse vision tasks. However, their deployment in federated settings is hindered by high computational demands, substantial communication overhead, and significant inference costs. We propose DSFedMed, a dual-scale federated framework that enables mutual knowledge distillation between a centralized foundation model and lightweight client models for medical image segmentation. To support knowledge distillation, a set of high-quality medical images is generated to replace real public datasets, and a learnability-guided sample selection strategy is proposed to enhance efficiency and effectiveness in dual-scale distillation. This mutual distillation enables the foundation model to transfer general knowledge to lightweight clients, while also incorporating client-specific insights to refine the foundation model. Evaluations on five medical imaging segmentation datasets show that DSFedMed achieves an average 2 percent improvement in Dice score while reducing communication costs and inference time by nearly 90 percent compared to existing federated foundation model baselines. These results demonstrate significant efficiency gains and scalability for resource-limited federated deployments.

cs.CV↗

arXiv2026

Minimum Star Partitions of Simple Polygons in Polynomial Time

We devise a polynomial-time algorithm for partitioning a simple polygon $P$ into a minimum number of star-shaped polygons. The question of whether such an algorithm exists has been open for more than four decades [Avis and Toussaint, Pattern Recognit., 1981] and it has been repeated frequently, for example in O'Rourke's famous book [Art Gallery Theorems and Algorithms, 1987]. In addition to its strong theoretical motivation, the problem is also motivated by practical domains such as CNC pocket milling, motion planning, and shape parameterization. The only previously known algorithm for a non-trivial special case is for $P$ being both monotone and rectilinear [Liu and Ntafos, Algorithmica, 1991]. For general polygons, an algorithm was only known for the restricted version in which Steiner points are disallowed [Keil, SIAM J. Comput., 1985], meaning that each corner of a piece in the partition must also be a corner of $P$. Interestingly, the solution size for the restricted version may be linear for instances where the unrestricted solution has constant size. The covering variant in which the pieces are star-shaped but allowed to overlap--known as the Art Gallery Problem--was recently shown to be $\exists\mathbb R$-complete and is thus likely not in NP [Abrahamsen, Adamaszek and Miltzow, STOC 2018 & J. ACM 2022]; this is in stark contrast to our result. Arguably the most related work to ours is the polynomial-time algorithm to partition a simple polygon into a minimum number of convex pieces by Chazelle and Dobkin [STOC, 1979 & Comp. Geom., 1985].

cs.CG↗

arXiv2026

Feature-Aware One-Shot Federated Learning via Hierarchical Token Sequences

One-shot federated learning (OSFL) reduces the communication cost and privacy risks of iterative federated learning by constructing a global model with a single round of communication. However, most existing methods struggle to achieve robust performance on real-world domains such as medical imaging, or are inefficient when handling non-IID (Independent and Identically Distributed) data. To address these limitations, we introduce FALCON, a framework that enhances the effectiveness of OSFL over non-IID image data. The core idea of FALCON is to leverage the feature-aware hierarchical token sequences generation and knowledge distillation into OSFL. First, each client leverages a pretrained visual encoder with hierarchical scale encoding to compress images into hierarchical token sequences, which capture multi-scale semantics. Second, a multi-scale autoregressive transformer generator is used to model the distribution of these token sequences and generate the synthetic sequences. Third, clients upload the synthetic sequences along with the local classifier trained on the real token sequences to the server. Finally, the server incorporates knowledge distillation into global training to reduce reliance on precise distribution modeling. Experiments on medical and natural image datasets validate the effectiveness of FALCON in diverse non-IID scenarios, outperforming the best OSFL baselines by 9.58% in average accuracy.

cs.LG↗

arXiv2025

Advancing Generative Artificial Intelligence and Large Language Models for Demand Side Management with Internet of Electric Vehicles

The energy optimization and demand side management (DSM) of Internet of Things (IoT)-enabled microgrids are being transformed by generative artificial intelligence, such as large language models (LLMs). This paper explores the integration of LLMs into energy management, and emphasizes their roles in automating the optimization of DSM strategies with Internet of Electric Vehicles (IoEV) as a representative example of the Internet of Vehicles (IoV). We investigate challenges and solutions associated with DSM and explore the new opportunities presented by leveraging LLMs. Then, we propose an innovative solution that enhances LLMs with retrieval-augmented generation for automatic problem formulation, code generation, and customizing optimization. The results demonstrate the effectiveness of our proposed solution in charging scheduling and optimization for electric vehicles, and highlight our solution's significant advancements in energy efficiency and user adaptability. This work shows LLMs' potential in energy optimization of the IoT-enabled microgrids and promotes intelligent DSM solutions.

cs.LG↗

arXiv2025

HiLoMix: Robust High- and Low-Frequency Graph Learning Framework for Mixing Address Association

As mixing services are increasingly being exploited by malicious actors for illicit transactions, mixing address association has emerged as a critical research task. A range of approaches have been explored, with graph-based models standing out for their ability to capture structural patterns in transaction networks. However, these approaches face two main challenges: label noise and label scarcity, leading to suboptimal performance and limited generalization. To address these, we propose HiLoMix, a graph-based learning framework specifically designed for mixing address association. First, we construct the Heterogeneous Attributed Mixing Interaction Graph (HAMIG) to enrich the topological structure. Second, we introduce frequency-aware graph contrastive learning that captures complementary structural signals from high- and low-frequency graph views. Third, we employ weak supervised learning that assigns confidence-based weighting to noisy labels. Then, we jointly train high-pass and low-pass GNNs using both unsupervised contrastive signals and confidence-based supervision to learn robust node representations. Finally, we adopt a stacking framework to fuse predictions from multiple heterogeneous models, further improving generalization and robustness. Experimental results demonstrate that HiLoMix outperforms existing methods in mixing address association.

cs.SI↗

arXiv2025

Solving the Correlation Cluster LP in Sublinear Time

Correlation Clustering is a fundamental and widely-studied problem in unsupervised learning and data mining. The input is a graph and the goal is to construct a clustering minimizing the number of inter-cluster edges plus the number of missing intra-cluster edges. CCL+24 introduced the cluster LP for Correlation Clustering, which they argued captures the problem much more succinctly than previous linear programming formulations. However, the cluster LP has exponential size, with a variable for every possible set of vertices in the input graph. Nevertheless, CCL+24 showed how to find a feasible solution for the cluster LP in time $O(n^{\text{poly}(1/ε)})$ with objective value at most $(1+ε)$ times the value of an optimal solution for the respective Correlation Clustering instance. Furthermore, they showed how to round a solution to the cluster LP, yielding a $(1.485+ε)$-approximation algorithm for the Correlation Clustering problem. The main technical result of this paper is a new approach to find a feasible solution for the cluster LP with objective value at most $(1+ε)$ of the optimum in time $\widetilde O(2^{\text{poly}(1/ε)} n)$, where $n$ is the number of vertices in the graph. We also show how to implement the rounding within the same time bounds, thus achieving a fast $(1.485+ε)$-approximation algorithm for the Correlation Clustering problem. This bridges the gap between state-of-the-art methods for approximating Correlation Clustering and the recent focus on fast algorithms.

cs.DS↗

arXiv2025

A highly accurate procedure for computing globally optimal Wannier functions in one-dimensional crystalline insulators

A standard task in solid state physics and quantum chemistry is the computation of localized molecular orbitals known as Wannier functions. In this manuscript, we propose a new procedure for computing Wannier functions in one-dimensional crystalline materials. Our approach proceeds by first performing parallel transport of the Bloch functions using numerical integration. Then a simple analytically computable correction is introduced to yield the optimally localized Wannier function. The resulting scheme is rapidly convergent and is proven to yield real-valued Wannier functions that achieve global optimality. The analysis in this manuscript can also be viewed as a proof of the existence of exponentially localized Wannier functions in one dimension. We illustrate the performance of the scheme by a number of numerical experiments.

math-ph↗

arXiv2025

Beyond Parameters: Exploring Virtual Logic Depth for Scaling Laws

Scaling large language models typically involves three dimensions: depth, width, and parameter count. In this work, we explore a fourth dimension, \textbf{virtual logical depth} (VLD), which increases effective algorithmic depth without changing parameter count by reusing weights. While parameter reuse is not new, its role in scaling has been underexplored. Unlike recent test-time methods that scale token-wise, VLD alters the internal computation graph during training and inference. Through controlled experiments, we obtain three key insights. (1) \textit{Knowledge capacity vs. parameters}: at fixed parameter count, VLD leaves knowledge capacity nearly unchanged, while across models capacity still scales with parameters. (2) \textit{Reasoning vs. reuse}: properly implemented VLD substantially improves reasoning ability \emph{without} more parameters, decoupling reasoning from size. This suggests a new scaling path beyond token-wise test-time methods. (3) \textit{Robustness and generality}: reasoning gains persist across architectures and reuse schedules, showing VLD captures a general scaling behavior. These results provide insight into future scaling strategies and raise a deeper question: does superintelligence require ever-larger models, or can it be achieved by reusing parameters and increasing logical depth? We argue many unknown dynamics in scaling remain to be explored. Code is available at https://anonymous.4open.science/r/virtual_logical_depth-8024/.

cs.AI↗

arXiv2025

Challenges and Trends in Egocentric Vision: A Survey

With the rapid development of artificial intelligence technologies and wearable devices, egocentric vision understanding has emerged as a new and challenging research direction, gradually attracting widespread attention from both academia and industry. Egocentric vision captures visual and multimodal data through cameras or sensors worn on the human body, offering a unique perspective that simulates human visual experiences. This paper provides a comprehensive survey of the research on egocentric vision understanding, systematically analyzing the components of egocentric scenes and categorizing the tasks into four main areas: subject understanding, object understanding, environment understanding, and hybrid understanding. We explore in detail the sub-tasks within each category. We also summarize the main challenges and trends currently existing in the field. Furthermore, this paper presents an overview of high-quality egocentric vision datasets, offering valuable resources for future research. By summarizing the latest advancements, we anticipate the broad applications of egocentric vision technologies in fields such as augmented reality, virtual reality, and embodied intelligence, and propose future research directions based on the latest developments in the field.

cs.CV↗

arXiv2025

Bounding a Polygon by a Minimum Number of Vertices

Suppose that a polygon $P$ is given as an array containing the vertices in counterclockwise order. We analyze how many vertices (including the index of each of these vertices) we need to know before we can bound $P$, i.e., report a bounded region $R$ in the plane such that $P\subset R$. We show that there exists polygons where $4\log_2 n+O(1)$ vertices are enough, while $\log_3n-o(\log n)$ must always be known. We thus answer the question up to a constant factor. This can be seen as an analysis of the shortest possible certificate or the best-case running time of any algorithm solving a variety of problems involving polygons, where a bound must be known in order to answer correctly. This includes various packing problems such as deciding whether a polygon can be contained inside another polygon.

cs.CG↗