arXiv subjects

Hao Zhu

Publications and source records attributed to Hao Zhu.

At least 55 records · Page 3Linked to original sources

arXiv2026

KOCO-BENCH: Can Large Language Models Leverage Domain Knowledge in Software Development?

Large language models (LLMs) excel at general programming but struggle with domain-specific software development, necessitating domain specialization methods for LLMs to learn and utilize domain knowledge and data. However, existing domain-specific code benchmarks cannot evaluate the effectiveness of domain specialization methods, which focus on assessing what knowledge LLMs possess rather than how they acquire and apply new knowledge, lacking explicit knowledge corpora for developing domain specialization methods. To this end, we present KOCO-BENCH, a novel benchmark designed for evaluating domain specialization methods in real-world software development. KOCO-BENCH contains 6 emerging domains with 11 software frameworks and 25 projects, featuring curated knowledge corpora alongside multi-granularity evaluation tasks including domain code generation (from function-level to project-level with rigorous test suites) and domain knowledge understanding (via multiple-choice Q&A). Unlike previous benchmarks that only provide test sets for direct evaluation, KOCO-BENCH requires acquiring and applying diverse domain knowledge (APIs, rules, constraints, etc.) from knowledge corpora to solve evaluation tasks. Our evaluations reveal that KOCO-BENCH poses significant challenges to state-of-the-art LLMs. Even with domain specialization methods (e.g., SFT, RAG, kNN-LM) applied, improvements remain marginal. Best-performing coding agent, Claude Code, achieves only 34.2%, highlighting the urgent need for more effective domain specialization methods. We release KOCO-BENCH, evaluation code, and baselines to advance further research at https://github.com/jiangxxxue/KOCO-bench.

cs.SE↗

arXiv2026

Learning the Weather-Grid Nexus via Weather-to-Voltage (W2V) Predictive Modeling

This paper proposes a weather-to-voltage (W2V) predictive modeling framework to learn the underlying weather-grid nexus. Unlike existing approaches on weather-informed grid operations, our proposed W2V model can achieve the joint analysis of weather and grid states, and further leverage this coupling to enhance grid-aware weather forecasting (GAWF) as a key application. To achieve this end-to-end learning, the W2V model acts as a differentiable surrogate for weather-incorporated power flow analysis by mapping weather features at high spatial resolution directly to grid-wide bus voltages. Thanks to a compact neural network design and principal component analysis based initialization, it achieves high voltage prediction accuracy and numerical stability during training. Building on this capability, W2V-based voltage signals are used to guide the development of GAWF that can account for its downstream voltage prediction performance. Using a 6717-bus Texas synthetic test system with meteorological inputs from 701 weather locations, our numerical tests have verified the excellent accuracy and generalizability of the proposed W2V model. More importantly, the W2V model has enabled the GAWF to effectively prioritize the weather features and conditions that are most critical to grid operations, such as system-wide quick wind drops preceding ramp-ups.

eess.SY↗

arXiv2026

Wide-Area Power System Oscillations from Large-Scale AI Workloads

This paper develops a new dynamic power profiling approach for modeling AI-centric datacenter loads and analyzing their impact on grid operations, particularly their potential to induce wide-area grid oscillations. We characterize the periodic stochastic power fluctuations inherent to large-scale AI workloads during both the training and fine-tuning stages, driven by the state-of-the-art graphics processing unit (GPU) computing architecture design. % and distributed mini-batch processing cycles. These sustained, large power fluctuations, unlike conventional load ramping, act as persistent forcing inputs capable of interacting with and amplifying local and inter-area oscillation modes. Using the WECC 179-bus system and the NPCC 140-bus system, we have numerically studied the amplitude and variability of oscillatory responses under different factors. These factors include system strength, penetration level, fluctuation frequency range, individual datacenter size, geographical deployment, fluctuation suppression level, and workload ratio. Simulation results show that, notably, narrower fluctuation bands, larger single-site capacities, or dispersed siting can intensify oscillations across multiple modes. Our models and numerical studies provide a quantitative basis for integrating AI-dominant electricity demand into grid oscillation studies and further support the development of new planning and operational measures to power the growth of AI/computing load demands.

eess.SY↗

arXiv2026

Hierarchically Robust Zero-shot Vision-language Models

Vision-Language Models (VLMs) can perform zero-shot classification but are susceptible to adversarial attacks. While robust fine-tuning improves their robustness, existing approaches align fixed text embeddings with an image embedding, sacrificing natural performance and robustness. A robustness degradation also occurs when a model faces adversarial attacks targeting superclasses (parent classes, e.g., mammal) in addition to their base (leaf) classes (e.g., cat). Thus, to enhance adversarial robustness and leverage the inherent hierarchical properties of class space, we propose a novel adversarial fine-tuning framework based on hierarchical embeddings and several levels of adversarially robust alignment of image-text modalities. Additional mechanisms place visual embeddings at the desired depth of hierarchy, and we provide a theoretical connection between the depth of embedding in the hierarchy and the maximum viable margin size. Our model naturally realizes several margin sizes, boosting generalization of adversaries for robustification. As various trees with different parent labels can share the same leaf labels, we also consider aligning over multiple trees to boost semantic variety. Experiments across several datasets are performed.

cs.CV↗

arXiv2026

TRASE-NODEs: Trajectory Sensitivity-aware Neural Ordinary Differential Equations for Efficient Dynamic Modeling

Modeling dynamical systems is crucial across the science and engineering fields for accurate prediction, control, and decision-making. Recently, machine learning (ML) approaches, particularly neural ordinary differential equations (NODEs), have emerged as a powerful tool for data-driven modeling of continuous-time dynamics. Nevertheless, standard NODEs require a large number of data samples to remain consistent under varying control inputs, posing challenges to generate sufficient simulated data and ensure the safety of control design. To address this gap, we propose trajectory-sensitivity-aware (TRASE-)NODEs, which construct an augmented system for both state and sensitivity, enabling simultaneous learning of their dynamics. This formulation allows the adjoint method to update gradients in a memory-efficient manner and ensures that time-invariant control set-point effects are captured in the learned dynamics. We evaluate TRASE-NODEs using damped oscillator and inverter-based resources (IBRs). The results show that TRASE-NODEs generalize better from the limited training data, yielding lower prediction errors than standard NODEs for both examples. The proposed framework offers a data-efficient, control-oriented modeling approach suitable for dynamic systems that require accurate trajectory sensitivity prediction.

eess.SY↗

arXiv2026

Giving Faces Their Feelings Back: Explicit Emotion Control for Feedforward Single-Image 3D Head Avatars

We present a framework for explicit emotion control in feed-forward, single-image 3D head avatar reconstruction. Unlike existing pipelines where emotion is implicitly entangled with geometry or appearance, we treat emotion as a first-class control signal that can be manipulated independently and consistently across identities. Our method injects emotion into existing feed-forward architectures via a dual-path modulation mechanism without modifying their core design. Geometry modulation performs emotion-conditioned normalization in the original parametric space, disentangling emotional state from speech-driven articulation, while appearance modulation captures identity-aware, emotion-dependent visual cues beyond geometry. To enable learning under this setting, we construct a time-synchronized, emotion-consistent multi-identity dataset by transferring aligned emotional dynamics across identities. Integrated into multiple state-of-the-art backbones, our framework preserves reconstruction and reenactment fidelity while enabling controllable emotion transfer, disentangled manipulation, and smooth emotion interpolation, advancing expressive and scalable 3D head avatars.

cs.CV↗

arXiv2026

LACE-S: Toward Sensitivity-consistent Locational Average Carbon Emissions via Neural Representation

Carbon-aware grid optimization relies on accurate locational emission metrics to effectively guide demand-side decarbonization tasks such as spatial load shifting. However, existing metrics are only valid around limited operating regions and unfortunately cannot generalize the emission patterns beyond these regions. When these metrics are used to signal carbon-sensitive resources, they could paradoxically increase system-wide emissions. This work seeks to develop a sensitivity-consistent metric for locational average carbon emissions (LACE-S) using a neural representation approach. To ensure physical validity, the neural model enforces total emission balance through an explicit projection layer while matching marginal emission sensitivities across the entire loading region. Jacobian-based regularization is further introduced to capture the underlying partition of load buses with closely aligned generator responses. Moreover, we present a scalable zonal aggregation strategy, ZACE-S, to reduce the model complexity by mapping nodal inputs to predefined market zones. Numerical tests on the IEEE 30-bus system have verified the performance improvements of LACE-S in matching total emissions and their sensitivities over the non-regularized design. Crucially, while spatial load shifting driven by existing metrics often increases the post-shift emissions, the proposed LACE-S metric has led to a reliable reduction of system-wide emissions, demonstrating its excellent consistency with the global emission patterns.

eess.SY↗

arXiv2026

Bringing Your Portrait to 3D Presence

We present a unified framework for reconstructing animatable 3D human avatars from a single portrait across head, half-body, and full-body inputs. Our method tackles three bottlenecks: pose- and framing-sensitive feature representations, limited scalable data, and unreliable proxy-mesh estimation. We introduce a Dual-UV representation that maps image features to a canonical UV space via Core-UV and Shell-UV branches, eliminating pose- and framing-induced token shifts. We also build a factorized synthetic data manifold combining 2D generative diversity with geometry-consistent 3D renderings, supported by a training scheme that improves realism and identity consistency. A robust proxy-mesh tracker maintains stability under partial visibility. Together, these components enable strong in-the-wild generalization. Trained only on half-body synthetic data, our model achieves state-of-the-art head and upper-body reconstruction and competitive full-body results. Extensive experiments and analyses further validate the effectiveness of our approach.

cs.CV↗

arXiv2026

UNICA: A Unified Neural Framework for Controllable 3D Avatars

Controllable 3D human avatars have found widespread applications in 3D games, the metaverse, and AR/VR scenarios. The conventional approach to creating such a 3D avatar requires a lengthy, intricate pipeline encompassing appearance modeling, motion planning, rigging, and physical simulation. In this paper, we introduce UNICA (UNIfied neural Controllable Avatar), a skeleton-free generative model that unifies all avatar control components into a single neural framework. Given keyboard inputs akin to video game controls, UNICA generates the next frame of a 3D avatar's geometry through an action-conditioned diffusion model operating on 2D position maps. A point transformer then maps the resulting geometry to 3D Gaussian Splatting for high-fidelity free-view rendering. Our approach naturally captures hair and loose clothing dynamics without manually designed physical simulation, and supports extra-long autoregressive generation. To the best of our knowledge, UNICA is the first model to unify the workflow of "motion planning, rigging, physical simulation, and rendering". Code is released at https://github.com/zjh21/UNICA.

cs.CV↗

arXiv2026

NeuroNarrator: A Generalist EEG-to-Text Foundation Model for Clinical Interpretation via Spectro-Spatial Grounding and Temporal State-Space Reasoning

Electroencephalography (EEG) provides a non-invasive window into neural dynamics at high temporal resolution and plays a pivotal role in clinical neuroscience research. Despite this potential, prevailing computational approaches to EEG analysis remain largely confined to task-specific classification objectives or coarse-grained pattern recognition, offering limited support for clinically meaningful interpretation. To address these limitations, we introduce NeuroNarrator, the first generalist EEG-to-text foundation model designed to translate electrophysiological segments into precise clinical narratives. A cornerstone of this framework is the curation of NeuroCorpus-160K, the first harmonized large-scale resource pairing over 160,000 EEG segments with structured, clinically grounded natural-language descriptions. Our architecture first aligns temporal EEG waveforms with spatial topographic maps via a rigorous contrastive objective, establishing spectro-spatially grounded representations. Building on this grounding, we condition a Large Language Model through a state-space-inspired formulation that integrates historical temporal and spectral context to support coherent clinical narrative generation. This approach establishes a principled bridge between continuous signal dynamics and discrete clinical language, enabling interpretable narrative generation that facilitates expert interpretation and supports clinical reporting workflows. Extensive evaluations across diverse benchmarks and zero-shot transfer tasks highlight NeuroNarrator's capacity to integrate temporal, spectral, and spatial dynamics, positioning it as a foundational framework for time-frequency-aware, open-ended clinical interpretation of electrophysiological data.

eess.SP↗

arXiv2026

Smoothing the Landscape: Causal Structure Learning via Diffusion Denoising Objectives

Understanding causal dependencies in observational data is critical for informing decision-making. These relationships are often modeled as Bayesian Networks (BNs) and Directed Acyclic Graphs (DAGs). Existing methods, such as NOTEARS and DAG-GNN, often face issues with scalability and stability in high-dimensional data, especially when there is a feature-sample imbalance. Here, we show that the denoising score matching objective of diffusion models could smooth the gradients for faster, more stable convergence. We also propose an adaptive k-hop acyclicity constraint that improves runtime over existing solutions that require matrix inversion. We name this framework Denoising Diffusion Causal Discovery (DDCD). Unlike generative diffusion models, DDCD utilizes the reverse denoising process to infer a parameterized causal structure rather than to generate data. We demonstrate the competitive performance of DDCDs on synthetic benchmarking data. We also show that our methods are practically useful by conducting qualitative analyses on two real-world examples. Code is available at this url: https://github.com/haozhu233/ddcd.

cs.LG↗

arXiv2026

VulInstruct: Teaching LLMs Root-Cause Reasoning for Vulnerability Detection via Security Specifications

Large language models (LLMs) have achieved remarkable progress in code understanding tasks. However, they demonstrate limited performance in vulnerability detection and struggle to distinguish vulnerable code from patched code. We argue that LLMs lack understanding of security specifications -- the expectations about how code should behave to remain safe. When code behavior differs from these expectations, it becomes a potential vulnerability. However, such knowledge is rarely explicit in training data, leaving models unable to reason about security flaws. We propose VulInstruct, a specification-guided approach that systematically extracts security specifications from historical vulnerabilities to detect new ones. VulInstruct constructs a specification knowledge base from two perspectives: (i) General specifications from high-quality patches across projects, capturing fundamental safe behaviors; and (ii) Domain-specific specifications from repeated violations in particular repositories relevant to the target code. VulInstruct retrieves relevant past cases and specifications, enabling LLMs to reason about expected safe behaviors rather than relying on surface patterns. We evaluate VulInstruct under strict criteria requiring both correct predictions and valid reasoning. On PrimeVul, VulInstruct achieves 45.0% F1-score (32.7% improvement) and 37.7% recall (50.8% improvement) compared to baselines, while uniquely detecting 24.3% of vulnerabilities -- 2.4x more than any baseline. In pair-wise evaluation, VulInstruct achieves 32.3% relative improvement. VulInstruct also discovered a previously unknown high-severity vulnerability (CVE-2025-56538) in production code, demonstrating practical value for real-world vulnerability discovery. All code and supplementary materials are available at https://github.com/zhuhaopku/VulInstruct-temp.

cs.SE↗

arXiv2026

LACON: Training Text-to-Image Model from Uncurated Data

The success of modern text-to-image generation is largely attributed to massive, high-quality datasets. Currently, these datasets are curated through a filter-first paradigm that aggressively discards low-quality raw data based on the assumption that it is detrimental to model performance. Is the discarded bad data truly useless, or does it hold untapped potential? In this work, we critically re-examine this question. We propose LACON (Labeling-and-Conditioning), a novel training framework that exploits the underlying uncurated data distribution. Instead of filtering, LACON re-purposes quality signals, such as aesthetic scores and watermark probabilities as explicit, quantitative condition labels. The generative model is then trained to learn the full spectrum of data quality, from bad to good. By learning the explicit boundary between high- and low-quality content, LACON achieves superior generation quality compared to baselines trained only on filtered data using the same compute budget, proving the significant value of uncurated data.

cs.CV↗

arXiv2026

Fitting Reinforcement Learning Model to Behavioral Data under Bandits

We consider the problem of fitting a reinforcement learning (RL) model to some given behavioral data under a multi-armed bandit environment. These models have received much attention in recent years for characterizing human and animal decision making behavior. We provide a generic mathematical optimization problem formulation for the fitting problem of a wide range of RL models that appear frequently in scientific research applications. We then provide a detailed theoretical analysis of its convexity properties. Based on the theoretical results, we introduce a novel solution method for the fitting problem of RL models based on convex relaxation and optimization. Our method is then evaluated in several simulated and real-world bandit environments to compare with some benchmark methods that appear in the literature. Numerical results indicate that our method achieves comparable performance to the state-of-the-art, while significantly reducing computation time. We also provide an open-source Python package for our proposed method to empower researchers to apply it in the analysis of their datasets directly, without prior knowledge of convex optimization.

cs.CE↗

arXiv2026

CrowdGaussian: Reconstructing High-Fidelity 3D Gaussians for Human Crowd from a Single Image

Single-view 3D human reconstruction has garnered significant attention in recent years. Despite numerous advancements, prior research has concentrated on reconstructing 3D models from clear, close-up images of individual subjects, often yielding subpar results in the more prevalent multi-person scenarios. Reconstructing 3D human crowd models is a highly intricate task, laden with challenges such as: 1) extensive occlusions, 2) low clarity, and 3) numerous and various appearances. To address this task, we propose CrowdGaussian, a unified framework that directly reconstructs multi-person 3D Gaussian Splatting (3DGS) representations from single-image inputs. To handle occlusions, we devise a self-supervised adaptation pipeline that enables the pretrained large human model to reconstruct complete 3D humans with plausible geometry and appearance from heavily occluded inputs. Furthermore, we introduce Self-Calibrated Learning (SCL). This training strategy enables single-step diffusion models to adaptively refine coarse renderings to optimal quality by blending identity-preserving samples with clean/corrupted image pairs. The outputs can be distilled back to enhance the quality of multi-person 3DGS representations. Extensive experiments demonstrate that CrowdGaussian generates photorealistic, geometrically coherent reconstructions of multi-person scenes.

cs.CV↗

arXiv2026

PGLib-CO2: A Power Grid Library for Real-Time Computation and Optimization of Carbon Emissions

Achieving a sustainable electricity infrastructure requires the explicit integration of carbon emissions into power system modeling and optimization. However, existing open-source test cases for power system research lack generator-level carbon profiling, preventing the benchmark of carbon-aware operational strategies. To address this gap, this work introduces PGLib-CO2, an open-source extension to the PGLib-OPF test case library. The proposed PGLib-CO2 enriches standard grid test cases with CO2 and CO2-equivalent emission intensity factors to achieve realistic, generator-level carbon profiling with an expanded list of fuel types. Using the standardized data, PGLib-CO2 allows us to enhance the algorithms for computing key carbon emission metrics. We first utilize the differentiable programming paradigm for computing LMCE by treating the OPF-based grid dispatch as a differentiable layer. This method provides a rigorous marginal sensitivity for general convex cost functions, eliminating the need of using a small incremental change in numerical perturbation. Moreover, to accelerate the real-time LMCE computation, we develop an MPP-based approach that shifts the optimization burden to offline phase of identifying the OPF critical regions. Since each critical region is characterized by a pre-computed affine dispatch function, the online phase reduces to identifying the region followed by efficiently evaluating the region-specific LMCE values. Numerical evaluations on IEEE test systems demonstrate that the differentiable LMCE computation attains the precise sensitivity information, and the MPP-based approach retrieves the LMCE signals faster than the direct optimization approach. By bridging high-fidelity data with advanced parametric computation, PGLib-CO2 provides a reproducible and computationally efficient foundation for future research in sustainable power system operations.

eess.SY↗

arXiv2026

SATURN: SAT-based Reinforcement Learning to Unleash LLMs Reasoning

How to design reinforcement learning (RL) tasks that effectively unleash the reasoning capability of large language models (LLMs) remains an open question. Existing RL tasks (e.g., math, programming, and constructing reasoning tasks) suffer from three key limitations: (1) Scalability. They rely heavily on human annotation or expensive LLM synthesis to generate sufficient training data. (2) Verifiability. LLMs' outputs are hard to verify automatically and reliably. (3) Controllable Difficulty. Most tasks lack fine-grained difficulty control, making it hard to train LLMs to develop reasoning ability from easy to hard. To address these limitations, we propose Saturn, a SAT-based RL framework that uses Boolean Satisfiability (SAT) problems to train and evaluate LLMs reasoning. Saturn enables scalable task construction, rule-based verification, and precise difficulty control. Saturn designs a curriculum learning pipeline that continuously improves LLMs' reasoning capability by constructing SAT tasks of increasing difficulty and training LLMs from easy to hard. To ensure stable training, we design a principled mechanism to control difficulty transitions. We introduce Saturn-2.6k, a dataset of 2,660 SAT problems with varying difficulty. It supports the evaluation of how LLM reasoning changes with problem difficulty. We apply Saturn to DeepSeek-R1-Distill-Qwen and obtain Saturn-1.5B and Saturn-7B. We achieve several notable results: (1) On SAT problems, Saturn-1.5B and Saturn-7B achieve average pass@3 improvements of +14.0 and +28.1, respectively. (2) On math and programming tasks, Saturn-1.5B and Saturn-7B improve average scores by +4.9 and +1.8 on benchmarks (e.g., AIME, LiveCodeBench). (3) Compared to the state-of-the-art (SOTA) approach in constructing RL tasks, Saturn achieves further improvements of +8.8%. We release the source code, data, and models to support future research.

cs.LG↗

arXiv2026

Carbon-aware Market Participation for Building Energy Management Systems

Tackling climate change requires the rapid and deep decarbonization of electric power systems. While energy management systems (EMSs) play a central role in this transition, conventional EMSs focus mainly on economic efficiency and often overlook the environmental impact of operational decisions. To address this gap, this paper proposes a unified, real-time building-level carbon-aware EMS (CAEMS) capable of simultaneously co-optimizing grid imports, energy storage, and flexible demand within a single integrated framework. We formulate a mixed-integer linear program (MILP) model that directly integrates time-varying marginal carbon intensity signals into the EMS objective for coordinated participation in both day-ahead (DA) and real-time (RT) markets. To relax the unrealistic assumption of perfect foresight, we incorporate a model predictive control (MPC) extension driven by a Transformer-based forecaster that jointly predicts electricity prices and carbon intensity. The proposed CAEMS is validated using real-world data from the PJM electricity market. Simulation results demonstrate that modest carbon prices can achieve a significant 22.5% reduction in emissions with only a 1.7% increase in cost.

eess.SY↗