arXiv subjects

Zixun Huang

Publications and source records attributed to Zixun Huang.

At least 19 recordsLinked to original sources

arXiv2026

FlowBalance: Verifier-Grounded Self-Improvement from On-Policy Reasoning Experience

A reasoning model can improve from its own on-policy experience, but this inner loop is fragile: terminal verifiers provide reliable yet sparse supervision, while dense same-model guidance can reinforce false confidence or overconcentrate learning on a narrow solution mode. We introduce FlowBalance, a verifier-grounded self-improvement method that learns a normalized distribution over complete responses. For each on-policy trajectory, a frozen training-time view of the same policy uses privileged context to produce token-level log-probability gains, which are aggregated into a trajectory-level self-guidance score. FlowBalance calibrates this score with the verifier-derived group advantage: guidance is retained on positive-advantage trajectories, reversed on negative-advantage trajectories, and disabled when the rollout group provides no outcome preference. The resulting energy exponentially reweights a reference policy, and profiled trajectory balance fits the normalized target with one log-partition estimate per rollout group. This realizes outcome-calibrated self-guidance via trajectory balance, without a separate token-level imitation loss. Our analysis establishes within-group contrast preservation, a minimum-change reverse-KL characterization, monotonic verifier control of target reward, and an exact correction against false-positive self-guidance on rejected responses. On mathematical reasoning, FlowBalance improves average performance over FlowRL on both Qwen3-4B and Qwen3-8B, while also improving training speed and stability, avoiding direct OPSD's response-length collapse, and exhibiting higher correct-strategy diversity in a controlled AIME24 diagnostic.

cs.LG↗

arXiv2026

Momentum in large-batch training: Polyak enlarges the critical batch size, Nesterov improves data efficiency

We study when and how momentum improves large-batch training in the one-pass regime, using power-law kernel regression as a tractable setting. We first characterize risk stability through the critical learning rate, defined as the largest learning rate for stable training, and obtain $η_{\mathrm{SGD}}^{\mathrm{crit}}\eqsim 1$, $η_{\mathrm{Polyak}}^{\mathrm{crit}}\eqsim \min\{1,B(1-ρ)\}$, and $η_{\mathrm{Nesterov}}^{\mathrm{crit}}\eqsim \min\{1,B^β(1-ρ)\}$, where $B$ is the batch size, $ρ$ is the momentum factor, and $β>1$ is the capacity exponent. Within this admissible region, we derive scaling laws for the full risk dynamics, capturing the progression from an early transient, through power-law decay, to a noise floor. We then minimize the final-step risk over the admissible learning rates and momentum factors under a fixed data budget, yielding a three-regime batch-size phase diagram that reveals how the role of momentum changes with batch size. Notably, Polyak enlarges the critical batch size, the largest batch size preserving the best small-batch data-scaling exponent, thereby enabling greater parallelism without sacrificing data efficiency. In contrast, Nesterov achieves better data efficiency in the large-batch regime because its look-ahead mechanism suppresses noise accumulation. Numerical experiments validate the predicted stability boundaries, risk dynamics, and batch-size phase diagram.

stat.ML↗

arXiv2026

TracingFlow: A Simulation-Free Trajectory Inference Framework Based on Second-Order Dynamics

Inferring continuous system evolution from sparse temporal snapshots is a key challenge in generative modeling and single-cell omics. While Optimal Transport (OT) is popular, existing frameworks are largely restricted to first-order dynamics, assuming memoryless velocity fields. This limits expressiveness, as first-order systems fail to account for regulatory momentum and time-delayed responses inherent in processes like cell differentiation. Here, we introduce TracingFlow, a simulation-free Flow Matching framework generalizing to second-order dynamics. By using neural networks to regress the acceleration field, TracingFlow provides an exact, efficient solution to the Dynamical Optimal Acceleration Transport (DOAT) problem. Unlike first-order methods yielding over-smoothed trajectories, our second-order formulation captures high-curvature transitions and nonlinear evolutions by learning the underlying force fields. Evaluated on complex synthetic and large-scale scRNA-seq datasets, TracingFlow achieves superior accuracy in distributional reconstruction and trajectory faithfulness. Moreover, by integrating lineage tracing priors, it recovers dynamical structures that are both mathematically optimal and biologically plausible.

cs.LG↗

arXiv2026

Recursive Synthesis for Long-Horizon Terminal Tasks

High-quality long-horizon training data for terminal agents is expensive to produce, often costing hundreds to thousands of dollars per task, because each task must keep the instruction, environment, reference solution, and verifier mutually consistent. Human authoring does not scale, and direct generation with large language models (LLMs) often breaks these dependencies. We present Recursive Synthetic Terminal Tasks (RST), a recursive verified synthesis framework for constructing long-horizon terminal-agent tasks at scale. Starting from verified seed tasks, RST extends the reference solution, realigns the verifier and instruction to the new workflow, validates the result in a fresh sandbox, and reuses accepted tasks as seeds for subsequent rounds. Across fifteen recursive rounds, RST produces 37,484 synthesized terminal-agent tasks at roughly $0.05 per task. Task difficulty increases substantially over rounds: the median reference solution grows from 67 to 374 lines, the median number of executed commands grows from 40 to 244, and DeepSeek-V4-Pro pass@4 drops from 90% at R1 to 2.5% at R15. To demonstrate training utility, we collect rejection-sampled Qwen3.5 trajectories on the synthesized tasks and use them for supervised fine-tuning. Fine-tuning on these trajectories improves Qwen3.5-27B and Qwen3.5-122B-A10B by up to 10 points on Terminal-Bench 2, Terminal-Bench Hard, and Long-Horizon Terminal Bench, while agentic PPO lifts Qwen3.5-27B to 49.44%, 32.00%, and 22.07% on the three benchmarks, corresponding to relative gains of 20.0%, 41.2%, and 21.9% over the base model. Moreover, after 15 rounds, the recursion shows no ceiling: synthesis yield and validation rates remain stable as difficulty keeps climbing, indicating that the process can continue well beyond the scale reported here.

cs.AI↗

arXiv2026

Variance-Aware Baselines and Adaptive Learning Rates for Reinforcement Learning with Verifiable Rewards

Reinforcement learning with verifiable rewards (RLVR) has emerged as an effective paradigm for post-training large language models, yet the design of its baselines and learning-rate schedules remains largely heuristic. This limits our understanding of the statistical properties of policy-gradient estimators and their interaction with optimization dynamics. In this work, we develop a theoretical framework for variance-aware baseline design and adaptive learning-rate selection in RLVR. Under a KL-regularized policy-optimization setting, we establish the unbiasedness of the resulting gradient estimator, derive exact variance expressions including the KL cross-covariance, and obtain an optimization-loss upper bound that enables principled reasoning about learning dynamics. Building on these results, we prove convergence guarantees and derive an adaptive learning-rate schedule governed by the signal-to-noise ratio (SNR) of the policy gradient. We further show that the variance-optimal baseline is a gradient-weighted estimator of the KL-regularized reward, providing a principled alternative to commonly used reward-based baselines. These results lead to two complementary improvements: a variance-optimal baseline and an SNR-adaptive learning-rate rule. Experiments on Qwen3-4B-Base show that each component independently improves policy-optimization performance. The learning-rate rule can also be naturally integrated with existing policy optimization methods to yield further gains, while combining it with the variance-optimal baseline gives the full Optimal Baseline and Learning-Rate Policy Optimization (OBLR-PO) method and achieves the strongest overall performance.

stat.ML↗

arXiv2026

NexusFlow: Unifying Disparate Tasks under Partial Supervision via Invertible Flow Networks

Partially Supervised Multi-Task Learning (PS-MTL) aims to leverage knowledge across tasks when annotations are incomplete. Existing approaches, however, have largely focused on the simpler setting of homogeneous, dense prediction tasks, leaving the more realistic challenge of learning from structurally diverse tasks unexplored. To this end, we introduce NexusFlow, a novel, lightweight, and plug-and-play framework effective in both settings. NexusFlow introduces a set of surrogate networks with invertible coupling layers to align the latent feature distributions of tasks, creating a unified representation that enables effective knowledge transfer. The coupling layers are bijective, preserving information while mapping features into a shared canonical space. This invertibility avoids representational collapse and enables alignment across structurally different tasks without reducing expressive capacity. We first evaluate NexusFlow on the core challenge of domain-partitioned autonomous driving, where dense map reconstruction and sparse multi-object tracking are supervised in different geographic regions, creating both structural disparity and a strong domain gap. NexusFlow sets a new state-of-the-art result on nuScenes, outperforming strong partially supervised baselines. To demonstrate generality, we further test NexusFlow on NYUv2 using three homogeneous dense prediction tasks, segmentation, depth, and surface normals, as a representative N-task PS-MTL scenario. NexusFlow yields consistent gains across all tasks, confirming its broad applicability.

cs.CV↗

arXiv2026

3DGEER: 3D Gaussian Rendering Made Exact and Efficient for Generic Cameras

3D Gaussian Splatting (3DGS) achieves an appealing balance between rendering quality and efficiency, but relies on approximating 3D Gaussians as 2D projections--an assumption that degrades accuracy, especially under generic large field-of-view (FoV) cameras. Despite recent extensions, no prior work has simultaneously achieved both projective exactness and real-time efficiency for general cameras. We introduce 3DGEER, a geometrically exact and efficient Gaussian rendering framework. From first principles, we derive a closed-form expression for integrating Gaussian density along a ray, enabling precise forward rendering and differentiable optimization under arbitrary camera models. To retain efficiency, we propose the Particle Bounding Frustum (PBF), which provides tight ray-Gaussian association without BVH traversal, and the Bipolar Equiangular Projection (BEAP), which unifies FoV representations, accelerates association, and improves reconstruction quality. Experiments on both pinhole and fisheye datasets show that 3DGEER outperforms prior methods across all metrics, runs 5x faster than existing projective exact ray-based baselines, and generalizes to wider FoVs unseen during training--establishing a new state of the art in real-time radiance field rendering.

cs.GR↗

arXiv2026

No Calibration, No Depth, No Problem: Cross-Sensor View Synthesis with 3D Consistency

We present the first study of cross-sensor view synthesis across different modalities. We examine a practical, fundamental, yet widely overlooked problem: getting aligned RGB-X data, where most RGB-X prior work assumes such pairs exist and focuses on modality fusion, but it empirically requires huge engineering effort in calibration. We propose a match-densify-consolidate method. First, we perform RGB-X image matching followed by guided point densification. Using the proposed confidence-aware densification and self-matching filtering, we attain better view synthesis and later consolidate them in 3D Gaussian Splatting (3DGS). Our method uses no 3D priors for X-sensor and only assumes nearly no-cost COLMAP for RGB. We aim to remove the cumbersome calibration for various RGB-X sensors and advance the popularity of cross-sensor learning by a scalable solution that breaks through the bottleneck in large-scale real-world RGB-X data collection.

cs.CV↗

arXiv2026

Functional Scaling Laws in Kernel Regression: Loss Dynamics and Learning Rate Schedules

Scaling laws have emerged as a unifying lens for understanding and guiding the training of large language models (LLMs). However, existing studies predominantly focus on the final-step loss, leaving open whether the entire loss dynamics obey similar laws and, crucially, how the learning rate schedule (LRS) shapes them. We address these gaps in a controlled theoretical setting by analyzing stochastic gradient descent (SGD) on a power-law kernel regression model. The key insight is a novel intrinsic-time viewpoint, which captures the training progress more faithfully than iteration count. We then establish a Functional Scaling Law (FSL) that captures the full loss trajectory under arbitrary LRSs, with the schedule's influence entering through a simple convolutional functional. We further instantiate the theory for three representative LRSs -- constant, exponential decay, and warmup-stable-decay (WSD) -- and derive explicit scaling relations in both data- and compute-limited regimes. These comparisons explain key empirical phenomena: (i) higher-capacity models are more data- and compute-efficient; (ii) learning-rate decay improves training efficiency; and (iii) WSD-type schedules outperform pure decay. Finally, experiments on LLMs ranging from 0.1B to 1B parameters demonstrate the practical relevance of FSL as a surrogate model for fitting and predicting loss trajectories in large-scale pre-training.

cs.LG↗

arXiv2025

SA$^{2}$Net: Scale-Adaptive Structure-Affinity Transformation for Spine Segmentation from Ultrasound Volume Projection Imaging

Spine segmentation, based on ultrasound volume projection imaging (VPI), plays a vital role for intelligent scoliosis diagnosis in clinical applications. However, this task faces several significant challenges. Firstly, the global contextual knowledge of spines may not be well-learned if we neglect the high spatial correlation of different bone features. Secondly, the spine bones contain rich structural knowledge regarding their shapes and positions, which deserves to be encoded into the segmentation process. To address these challenges, we propose a novel scale-adaptive structure-aware network (SA$^{2}$Net) for effective spine segmentation. First, we propose a scale-adaptive complementary strategy to learn the cross-dimensional long-distance correlation features for spinal images. Second, motivated by the consistency between multi-head self-attention in Transformers and semantic level affinity, we propose structure-affinity transformation to transform semantic features with class-specific affinity and combine it with a Transformer decoder for structure-aware reasoning. In addition, we adopt a feature mixing loss aggregation method to enhance model training. This method improves the robustness and accuracy of the segmentation process. The experimental results demonstrate that our SA$^{2}$Net achieves superior segmentation performance compared to other state-of-the-art methods. Moreover, the adaptability of SA$^{2}$Net to various backbones enhances its potential as a promising tool for advanced scoliosis diagnosis using intelligent spinal image analysis. The code and experimental demo are available at https://github.com/taetiseo09/SA2Net.

cs.CV↗

arXiv2025

Towards Generalizable Context-aware Anomaly Detection: A Large-scale Benchmark in Cloud Environments

Anomaly detection in cloud environments remains both critical and challenging. Existing context-level benchmarks typically focus on either metrics or logs and often lack reliable annotation, while most detection methods emphasize point anomalies within a single modality, overlooking contextual signals and limiting real-world applicability. Constructing a benchmark for context anomalies that combines metrics and logs is inherently difficult: reproducing anomalous scenarios on real servers is often infeasible or potentially harmful, while generating synthetic data introduces the additional challenge of maintaining cross-modal consistency. We introduce CloudAnoBench, a large-scale benchmark for context anomalies in cloud environments, comprising 28 anomalous scenarios and 16 deceptive normal scenarios, with 1,252 labeled cases and roughly 200,000 log and metric entries. Compared with prior benchmarks, CloudAnoBench exhibits higher ambiguity and greater difficulty, on which both prior machine learning methods and vanilla LLM prompting perform poorly. To demonstrate its utility, we further propose CloudAnoAgent, an LLM-based agent enhanced by symbolic verification that integrates metrics and logs. This agent system achieves substantial improvements in both anomaly detection and scenario identification on CloudAnoBench, and shows strong generalization to existing datasets. Together, CloudAnoBench and CloudAnoAgent lay the groundwork for advancing context-aware anomaly detection in cloud systems. Project Page: https://jayzou3773.github.io/cloudanobench-agent/

cs.AI↗

arXiv2025

Jenga Stacking Based on 6D Pose Estimation for Architectural Form Finding Process

This paper includes a review of current state of the art 6d pose estimation methods, as well as a discussion of which pose estimation method should be used in two types of architectural design scenarios. Taking the latest pose estimation research Gen6d as an example, we make a qualitative assessment of the current openset methods in terms of application level, prediction speed, resistance to occlusion, accuracy, resistance to environmental interference, etc. In addition, we try to combine 6D pose estimation and building wind environment assessment to create tangible architectural design approach, we discuss the limitations of the method and point out the direction in which 6d pose estimation is eager to progress in this scenario.

cs.CV↗

arXiv2025

2024 NASA SUITS Report: LLM-Driven Immersive Augmented Reality User Interface for Robotics and Space Exploration

As modern computing advances, new interaction paradigms have emerged, particularly in Augmented Reality (AR), which overlays virtual interfaces onto physical objects. This evolution poses challenges in machine perception, especially for tasks like 3D object pose estimation in complex, dynamic environments. Our project addresses critical issues in human-robot interaction within mobile AR, focusing on non-intrusive, spatially aware interfaces. We present URSA, an LLM-driven immersive AR system developed for NASA's 2023-2024 SUITS challenge, targeting future spaceflight needs such as the Artemis missions. URSA integrates three core technologies: a head-mounted AR device (e.g., HoloLens) for intuitive visual feedback, voice control powered by large language models for hands-free interaction, and robot tracking algorithms that enable accurate 3D localization in dynamic settings. To enhance precision, we leverage digital twin localization technologies, using datasets like DTTD-Mobile and specialized hardware such as the ZED2 camera for real-world tracking under noise and occlusion. Our system enables real-time robot control and monitoring via an AR interface, even in the absence of ground-truth sensors--vital for hazardous or remote operations. Key contributions include: (1) a non-intrusive AR interface with LLM-based voice input; (2) a ZED2-based dataset tailored for non-rigid robotic bodies; (3) a Local Mission Control Console (LMCC) for mission visualization; (4) a transformer-based 6DoF pose estimator (DTTDNet) optimized for depth fusion and real-time tracking; and (5) end-to-end integration for astronaut mission support. This work advances digital twin applications in robotics, offering scalable solutions for both aerospace and industrial domains.

cs.RO↗

arXiv2025

Robust 6DoF Pose Estimation Against Depth Noise and a Comprehensive Evaluation on a Mobile Dataset

Robust 6DoF pose estimation with mobile devices is the foundation for applications in robotics, augmented reality, and digital twin localization. In this paper, we extensively investigate the robustness of existing RGBD-based 6DoF pose estimation methods against varying levels of depth sensor noise. We highlight that existing 6DoF pose estimation methods suffer significant performance discrepancies due to depth measurement inaccuracies. In response to the robustness issue, we present a simple and effective transformer-based 6DoF pose estimation approach called DTTDNet, featuring a novel geometric feature filtering module and a Chamfer distance loss for training. Moreover, we advance the field of robust 6DoF pose estimation and introduce a new dataset -- Digital Twin Tracking Dataset Mobile (DTTD-Mobile), tailored for digital twin object tracking with noisy depth data from the mobile RGBD sensor suite of the Apple iPhone 14 Pro. Extensive experiments demonstrate that DTTDNet significantly outperforms state-of-the-art methods at least 4.32, up to 60.74 points in ADD metrics on the DTTD-Mobile. More importantly, our approach exhibits superior robustness to varying levels of measurement noise, setting a new benchmark for robustness to measurement noise. The project page is publicly available at https://openark-berkeley.github.io/DTTDNet/.

cs.CV↗

arXiv2024

Morphological-consistent Diffusion Network for Ultrasound Coronal Image Enhancement

Ultrasound curve angle (UCA) measurement provides a radiation-free and reliable evaluation for scoliosis based on ultrasound imaging. However, degraded image quality, especially in difficult-to-image patients, can prevent clinical experts from making confident measurements, even leading to misdiagnosis. In this paper, we propose a multi-stage image enhancement framework that models high-quality image distribution via a diffusion-based model. Specifically, we integrate the underlying morphological information from images taken at different depths of the 3D volume to calibrate the reverse process toward high-quality and high-fidelity image generation. This is achieved through a fusion operation with a learnable tuner module that learns the multi-to-one mapping from multi-depth to high-quality images. Moreover, the separate learning of the high-quality image distribution and the spinal features guarantees the preservation of consistent spinal pose descriptions in the generated images, which is crucial in evaluating spinal deformities. Remarkably, our proposed enhancement algorithm significantly outperforms other enhancement-based methods on ultrasound images in terms of image quality. Ultimately, we conduct the intra-rater and inter-rater measurements of UCA and higher ICC (0.91 and 0.89 for thoracic and lumbar angles) on enhanced images, indicating our method facilitates the measurement of ultrasound curve angles and offers promising prospects for automated scoliosis diagnosis.

eess.IV↗

arXiv2024

Encoding Urban Ecologies: Automated Building Archetype Generation through Self-Supervised Learning for Energy Modeling

As the global population and urbanization expand, the building sector has emerged as the predominant energy consumer and carbon emission contributor. The need for innovative Urban Building Energy Modeling grows, yet existing building archetypes often fail to capture the unique attributes of local buildings and the nuanced distinctions between different cities, jeopardizing the precision of energy modeling. This paper presents an alternative tool employing self-supervised learning to distill complex geometric data into representative, locale-specific archetypes. This study attempts to foster a new paradigm of interaction with built environments, incorporating local parameters to conduct bespoke energy simulations at the community level. The catered archetypes can augment the precision and applicability of energy consumption modeling at different scales across diverse building inventories. This tool provides a potential solution that encourages the exploration of emerging local ecologies. By integrating building envelope characteristics and cultural granularity into the building archetype generation process, we seek a future where architecture and urban design are intricately interwoven with the energy sector in shaping our built environments.

cs.CV↗

arXiv2023

Can Machine Learning Uncover Insights into Vehicle Travel Demand from Our Built Environment?

In this paper, we propose a machine learning-based approach to address the lack of ability for designers to optimize urban land use planning from the perspective of vehicle travel demand. Research shows that our computational model can help designers quickly obtain feedback on the vehicle travel demand, which includes its total amount and temporal distribution based on the urban function distribution designed by the designers. It also assists in design optimization and evaluation of the urban function distribution from the perspective of vehicle travel. We obtain the city function distribution information and vehicle hours traveled (VHT) information by collecting the city point-of-interest (POI) data and online vehicle data. The artificial neural networks (ANNs) with the best performance in prediction are selected. By using data sets collected in different regions for mutual prediction and remapping the predictions onto a map for visualization, we evaluate the extent to which the computational model sees use across regions in an attempt to reduce the workload of future urban researchers. Finally, we demonstrate the application of the computational model to help designers obtain feedback on vehicle travel demand in the built environment and combine it with genetic algorithms to optimize the current state of the urban environment to provide recommendations to designers.

cs.LG↗

arXiv2023

MARL: Multi-scale Archetype Representation Learning for Urban Building Energy Modeling

Building archetypes, representative models of building stock, are crucial for precise energy simulations in Urban Building Energy Modeling. The current widely adopted building archetypes are developed on a nationwide scale, potentially neglecting the impact of local buildings' geometric specificities. We present Multi-scale Archetype Representation Learning (MARL), an approach that leverages representation learning to extract geometric features from a specific building stock. Built upon VQ-AE, MARL encodes building footprints and purifies geometric information into latent vectors constrained by multiple architectural downstream tasks. These tailored representations are proven valuable for further clustering and building energy modeling. The advantages of our algorithm are its adaptability with respect to the different building footprint sizes, the ability for automatic generation across multi-scale regions, and the preservation of geometric features across neighborhoods and local ecologies. In our study spanning five regions in LA County, we show MARL surpasses both conventional and VQ-AE extracted archetypes in performance. Results demonstrate that geometric feature embeddings significantly improve the accuracy and reliability of energy consumption estimates. Code, dataset and trained models are publicly available: https://github.com/ZixunHuang1997/MARL-BuildingEnergyEstimation

cs.LG↗