arXiv subjects

Xuan Zhang

Publications and source records attributed to Xuan Zhang.

At least 55 records · Page 3Linked to original sources

arXiv2026

Geoparsing: Diagram Parsing for Plane and Solid Geometry with a Unified Formal Language

Multimodal Large Language Models (MLLMs) have achieved remarkable progress but continue to struggle with geometric reasoning, primarily due to the perception bottleneck regarding fine-grained visual elements. While formal languages have aided plane geometry understanding, solid geometry which requires spatial understanding remains largely unexplored. In this paper, we address this challenge by designing a unified formal language that integrates plane and solid geometry, comprehensively covering geometric structures and semantic relations. We construct GDP-29K, a large-scale dataset comprising 20k plane and 9k solid geometry samples collected from diverse real-world sources, each paired with its ground-truth formal description. To ensure syntactic correctness and geometric consistency, we propose a training paradigm that combines Supervised Fine-Tuning with Reinforcement Learning via Verifiable Rewards. Experiments show that our approach achieves state-of-the-art parsing performance. Furthermore, we demonstrate that our parsed formal descriptions serve as a critical cognitive scaffold, significantly boosting MLLMs' capabilities for downstream geometry reasoning tasks. Our data and code are available at Geoparsing.

cs.CV↗

arXiv2026

Mean dimension of general iterated function systems

In this paper, we introduce and investigate the notions of Mean Dimension and Metric Mean Dimension for generalized iterated function systems (IFS). We establish basic properties of these invariants and prove that Mean Dimension is always bounded above by the lower Metric Mean Dimension and the upper Metric Mean Dimension in this setting. We further show that generalized iterated function systems with the Small Boundary Property have zero Mean Dimension. Finally, we introduce a Gluing Orbit Property for generalized iterated function systems and prove that, under suitable transitivity and non-rigidity assumptions, it guarantees positive topological entropy.

math.DS↗

arXiv2026

The Eleventh NTIRE 2026 Efficient Super-Resolution Challenge Report

This paper reviews the NTIRE 2026 challenge on efficient single-image super-resolution with a focus on the proposed solutions and results. The aim of this challenge is to devise a network that reduces one or several aspects, such as runtime, parameters, and FLOPs, while maintaining PSNR of around 26.90 dB on the DIV2K_LSDIR_valid dataset, and 26.99 dB on the DIV2K_LSDIR_test dataset. The challenge had 95 registered participants, and 15 teams made valid submissions. They gauge the state-of-the-art results for efficient single-image super-resolution.

cs.CV↗

arXiv2026

PowerGenie: Analytically-Guided Evolutionary Discovery of Superior Reconfigurable Power Converters

Discovering superior circuit topologies requires navigating an exponentially large design space-a challenge traditionally reserved for human experts. Existing AI methods either select from predefined templates or generate novel topologies at a limited scale without rigorous verification, leaving large-scale performance-driven discovery underexplored. We present PowerGenie, a framework for automated discovery of higher-performance reconfigurable power converters at scale. PowerGenie introduces: (1) an automated analytical framework that determines converter functionality and theoretical performance limits without component sizing or SPICE simulation, and (2) an evolutionary finetuning method that co-evolves a generative model with its training distribution through fitness selection and uniqueness verification. Unlike existing methods that suffer from mode collapse and overfitting, our approach achieves higher syntax validity, function validity, novelty rate, and figure-of-merit (FoM). PowerGenie discovers a novel 8-mode reconfigurable converter with 23% higher FoM than the best training topology. SPICE simulations confirm average absolute efficiency gains of 10% across 8 modes and up to 17% at a single mode. Code will be released upon publication.

cs.LG↗

arXiv2026

Droplet impact on a superhydrophobic surface under shear airflow: Lattice Boltzmann simulations and scaling analyses

Droplet impact in airflow environments is ubiquitous in nature and industry, making the understanding of this multiphase behavior crucial for technologies such as anti-icing and spray cooling. In this study, the dynamics of droplet impact on a superhydrophobic surface under shear airflow are numerically investigated using the pseudopotential multiphase lattice Boltzmann method. This three-dimensional model employs a non-orthogonal multiple-relaxation-time scheme to enhance numerical stability and a contact angle hysteresis window to effectively capture dynamic wetting. Specifically, the kinetic energy supplied by the airflow enhances streamwise spreading and significantly expands the final contact footprint due to continuous horizontal sliding. To describe the nonlinear dependence of these contact-line characteristics on the impact Weber number (We) and the airflow Reynolds number (Re), a set of composite scaling laws is developed based on a modified Weber number (We*) that incorporates the airflow contribution. Moreover, the aerodynamic effect leads to a higher velocity restitution coefficient and a deflected take-off angle. Based on an energy partition analysis at detachment, a refined power law is derived to scale the vertical restitution coefficient under shear airflow, while the streamwise restitution coefficient is formulated via the sliding velocity approximation. Integrating these two directional components enables accurate quantitative predictions of the total restitution coefficient and the take-off angle governed by the interplay of We and Re. Overall, this study clarifies the underlying mechanisms of droplet-airflow-surface interactions, providing practical insights for predicting droplet behaviors and guiding surface design under aerodynamic conditions.

physics.flu-dyn↗

arXiv2026

CircuitSense: A Hierarchical MLLM Benchmark Bridging Visual Comprehension and Symbolic Reasoning in Engineering Design Process

Engineering design operates through hierarchical abstraction from system specifications to component implementations, requiring visual understanding coupled with mathematical reasoning at each level. While Multi-modal Large Language Models (MLLMs) excel at natural image tasks, their ability to extract mathematical models from technical diagrams remains unexplored. We present \textbf{CircuitSense}, a comprehensive benchmark evaluating circuit understanding across this hierarchy through 8,006+ problems spanning component-level schematics to system-level block diagrams. Our benchmark uniquely examines the complete engineering workflow: Perception, Analysis, and Design, with a particular emphasis on the critical but underexplored capability of deriving symbolic equations from visual inputs. We introduce a hierarchical synthetic generation pipeline consisting of a grid-based schematic generator and a block diagram generator with auto-derived symbolic equation labels. Comprehensive evaluation of six state-of-the-art MLLMs, including both closed-source and open-source models, reveals fundamental limitations in visual-to-mathematical reasoning. Closed-source models achieve over 85\% accuracy on perception tasks involving component recognition and topology identification, yet their performance on symbolic derivation and analytical reasoning falls below 19\%, exposing a critical gap between visual parsing and symbolic reasoning. Models with stronger symbolic reasoning capabilities consistently achieve higher design task accuracy, confirming the fundamental role of mathematical understanding in circuit synthesis and establishing symbolic reasoning as the key metric for engineering competence.

cs.CV↗

arXiv2026

Echo: Towards Advanced Audio Comprehension via Audio-Interleaved Reasoning

The maturation of Large Audio Language Models (LALMs) has raised growing expectations for them to comprehend complex audio much like humans. Current efforts primarily replicate text-based reasoning by contextualizing audio content through a one-time encoding, which introduces a critical information bottleneck. Drawing inspiration from human cognition, we propose audio-interleaved reasoning to break through this bottleneck. It treats audio as an active reasoning component, enabling sustained audio engagement and perception-grounded analysis. To instantiate it, we introduce a two-stage training framework, first teaching LALMs to localize salient audio segments through supervised fine-tuning, and then incentivizing proficient re-listening via reinforcement learning. In parallel, a structured data generation pipeline is developed to produce high-quality training data. Consequently, we present Echo, a LALM capable of dynamically re-listening to audio in demand during reasoning. On audio comprehension benchmarks, Echo achieves overall superiority in both challenging expert-level and general-purpose tasks. Comprehensive analysis further confirms the efficiency and generalizability of audio-interleaved reasoning, establishing it as a promising direction for advancing audio comprehension. Project page: https://github.com/wdqqdw/Echo.

cs.SD↗

arXiv2026

Learning to Evolve: Bayesian-Guided Continual Knowledge Graph Embedding

As social media and the World Wide Web become hubs for information dissemination, effectively organizing and understanding the vast amounts of dynamically evolving Web content is crucial. Knowledge graphs (KGs) provide a powerful framework for structuring this information. However, the rapid emergence of new hot topics, user relationships, and events in social media renders traditional static knowledge graph embedding (KGE) models rapidly outdated. Continual Knowledge Graph Embedding (CKGE) aims to address this issue, but existing methods commonly suffer from catastrophic forgetting, whereby older, but still valuable, information is lost when learning new knowledge (such as new memes or trending events). This means the model cannot effectively learn the evolution of the data. We propose a novel CKGE framework, BAKE. Unlike existing methods, BAKE formulates CKGE as a sequential Bayesian inference problem and utilizes the Bayesian posterior update principle as a natural continual learning strategy. This principle is insensitive to data order and provides theoretical guarantees to preserve prior knowledge as much as possible. Specifically, we treat each batch of new data as a Bayesian update to the model's prior. By maintaining the posterior distribution, the model effectively preserves earlier knowledge even as it evolves over multiple snapshots. Furthermore, to constrain the evolution of knowledge across snapshots, we introduce a continual clustering method that maintains the compact cluster structure of entity embeddings through a regularization term, ensuring semantic consistency while allowing controlled adaptation to new knowledge. We conduct extensive experiments on multiple CKGE benchmarks, which demonstrate that BAKE achieves the top performance in the vast majority of cases compared to existing approaches.

cs.CL↗

arXiv2026

Synergy over Discrepancy: A Partition-Based Approach to Multi-Domain LLM Fine-Tuning

Large language models (LLMs) demonstrate impressive generalization abilities, yet adapting them effectively across multiple heterogeneous domains remains challenging due to inter-domain interference. To overcome this challenge, we propose a partition-based multi-stage fine-tuning framework designed to exploit inter-domain synergies while minimizing negative transfer. Our approach strategically partitions domains into subsets (stages) by balancing domain discrepancy, synergy, and model capacity constraints. We theoretically analyze the proposed framework and derive novel generalization bounds that justify our partitioning strategy. Extensive empirical evaluations on various language understanding tasks show that our method consistently outperforms state-of-the-art baselines.

cs.LG↗

arXiv2026

Aligning by Misaligning: Boundary-aware Curriculum Learning for Multimodal Alignment

Most multimodal models treat every negative pair alike, ignoring the ambiguous negatives that differ from the positive by only a small detail. We propose Boundary-Aware Curriculum with Local Attention (BACL), a lightweight add-on that turns these borderline cases into a curriculum signal. A Boundary-aware Negative Sampler gradually raises difficulty, while a Contrastive Local Attention loss highlights where the mismatch occurs. The two modules are fully differentiable and work with any off-the-shelf dual encoder. Theory predicts a fast O(1/n) error rate; practice shows up to +32% R@1 over CLIP and new SOTA on four large-scale benchmarks, all without extra labels.

cs.LG↗

arXiv2026

First Submillimeter Lights from Dome A: Tracing the Carbon Cycle in the Feedback of Massive Stars

The cycling of carbon between its ionized, atomic, and molecular phases shapes the chemical compositions and physical conditions of the interstellar medium (ISM). However, ground-based studies of the full carbon cycle have been limited by atmospheric absorption. Dome~A, the most promising site for submillimeter astronomy, has long resisted successful submillimeter astronomical observations. Using the 60~cm Antarctic Terahertz Explorer, we present the first successful CO ($4-3$) and [CI] ($^3P_1 - ^3P_0$) mapping observations of two archetypal triggered massive star-formation regions at Dome~A. These data, together with archival [CII], provide the first complete characterization of all three carbon phases in these environments. We find elevated C$^{0}$/CO abundance ratios in high-extinction regions, plausibly driven by deep penetration of intense radiation fields from massive stars into a clumpy ISM. These findings mark a major milestone for submillimeter astronomy at Dome~A and offer valuable insights into the impact of massive star feedback on the surrounding ISM.

astro-ph.GA↗

arXiv2026

1-GHz VIS-to-MIR frequency combs enabled by CMOS-compatible nanophotonic waveguides

A fully stabilized frequency comb is essential for precision metrology and coherent optical synthesis. However, fully-stabilized frequency combs generally require separate stages for supercontinuum generation (SCG) and self-referencing, largely limiting their compactness. Here, enabled by the low-threshold multi-octave supercontinuum generation and concurrent third-harmonic generation in low-loss silicon nitride waveguides, we present a novel approach to a self-referenced frequency comb source at 1 GHz repetition rate spanning from the full visible (VIS) to the mid-infrared (MIR). Our coherent comb is seeded by an all-polarization-maintaining ultrafast fiber laser at 1556 nm, with a pulse duration of 73 fs at 1 GHz repetition rate. With an injected energy of merely 110 pJ, the pulses propagate through dispersion-engineered Si3N4 waveguides, generating supercontinuum spanning over three octaves from 350-3280 nm i.e. 0.76 PHz of coherent bandwidth. Moreover, the on-chip third harmonic generation provides a carrier envelope offset beat note via f-3f with a signal-to-noise ratio of 43 dB. Fueled by the evolving photonic integration providing possibilities of on-chip filtering and photo-detectors, this approach for single-chip self-referencing of high-repetition-rate frequency combs paves the way for ultrabroadband comb sources with unprecedented compactness and field-readiness.

physics.optics↗

arXiv2025

AnalogSAGE: Self-evolving Analog Design Multi-Agents with Stratified Memory and Grounded Experience

Analog circuit design remains a knowledge- and experience-intensive process that relies heavily on human intuition for topology generation and device parameter tuning. Existing LLM-based approaches typically depend on prompt-driven netlist generation or predefined topology templates, limiting their ability to satisfy complex specification requirements. We propose AnalogSAGE, an open-source self-evolving multi-agent framework that coordinates three-stage agent explorations through four stratified memory layers, enabling iterative refinement with simulation-grounded feedback. To support reproducibility and generality, we release the source code. Our benchmark spans ten specification-driven operational amplifier design problems of varying difficulty, enabling quantitative and cross-task comparison under identical conditions. Evaluated under the open-source SKY130 PDK with ngspice, AnalogSAGE achieves a 10$\times$ overall pass rate, a 48$\times$ Pass@1, and a 4$\times$ reduction in parameter search space compared with existing frameworks, demonstrating that stratified memory and grounded reasoning substantially enhance the reliability and autonomy of analog design automation in practice.

cs.AR↗

arXiv2025

Learn the Ropes, Then Trust the Wins: Self-imitation with Progressive Exploration for Agentic Reinforcement Learning

Reinforcement learning (RL) is the dominant paradigm for sharpening strategic tool use capabilities of LLMs on long-horizon, sparsely-rewarded agent tasks, yet it faces a fundamental challenge of exploration-exploitation trade-off. Existing studies stimulate exploration through the lens of policy entropy, but such mechanical entropy maximization is prone to RL instability due to the multi-turn distribution shifting. In this paper, we target the progressive exploration-exploitation balance under the guidance of the agent's own experiences without succumbing to either entropy collapsing or runaway divergence. We propose SPEAR, a self-imitation learning (SIL) recipe for training agentic LLMs. It extends the vanilla SIL, where a replay buffer stores good experience for off-policy update, by gradually steering the policy entropy across stages. Specifically, the proposed curriculum scheduling harmonizes intrinsic reward shaping and self-imitation to 1) expedite exploration via frequent tool interactions at the beginning, and 2) strengthen exploitation of successful tactics upon convergence towards familiarity with the environment. We also combine bag-of-tricks of industrial RL optimizations for a strong baseline Dr.BoT to demonstrate our effectiveness. In ALFWorld and WebShop, SPEAR increases the success rates of GRPO/GiGPO/Dr.BoT by up to 16.1%/5.1%/8.6% and 20.7%/11.8%/13.9%, respectively. In AIME24 and AIME25, SPEAR boosts Dr.BoT by up to 3.8% and 6.1%, respectively. Such gains incur only 10%-25% extra theoretical complexity and negligible runtime overhead in practice, demonstrating the plug-and-play scalability of SPEAR.

cs.LG↗

arXiv2025

BanditSpec: Adaptive Speculative Decoding via Bandit Algorithms

Speculative decoding has emerged as a popular method to accelerate the inference of Large Language Models (LLMs) while retaining their superior text generation performance. Previous methods either adopt a fixed speculative decoding configuration regardless of the prefix tokens, or train draft models in an offline or online manner to align them with the context. This paper proposes a training-free online learning framework to adaptively choose the configuration of the hyperparameters for speculative decoding as text is being generated. We first formulate this hyperparameter selection problem as a Multi-Armed Bandit problem and provide a general speculative decoding framework BanditSpec. Furthermore, two bandit-based hyperparameter selection algorithms, UCBSpec and EXP3Spec, are designed and analyzed in terms of a novel quantity, the stopping time regret. We upper bound this regret under both stochastic and adversarial reward settings. By deriving an information-theoretic impossibility result, it is shown that the regret performance of UCBSpec is optimal up to universal constants. Finally, extensive empirical experiments with LLaMA3 and Qwen2 demonstrate that our algorithms are effective compared to existing methods, and the throughput is close to the oracle best hyperparameter in simulated real-life LLM serving scenarios with diverse input prompts.

cs.LG↗

arXiv2025

DeepKnown-Guard: A Proprietary Model-Based Safety Response Framework for AI Agents

With the widespread application of Large Language Models (LLMs), their associated security issues have become increasingly prominent, severely constraining their trustworthy deployment in critical domains. This paper proposes a novel safety response framework designed to systematically safeguard LLMs at both the input and output levels. At the input level, the framework employs a supervised fine-tuning-based safety classification model. Through a fine-grained four-tier taxonomy (Safe, Unsafe, Conditionally Safe, Focused Attention), it performs precise risk identification and differentiated handling of user queries, significantly enhancing risk coverage and business scenario adaptability, and achieving a risk recall rate of 99.3%. At the output level, the framework integrates Retrieval-Augmented Generation (RAG) with a specifically fine-tuned interpretation model, ensuring all responses are grounded in a real-time, trustworthy knowledge base. This approach eliminates information fabrication and enables result traceability. Experimental results demonstrate that our proposed safety control model achieves a significantly higher safety score on public safety evaluation benchmarks compared to the baseline model, TinyR1-Safety-8B. Furthermore, on our proprietary high-risk test set, the framework's components attained a perfect 100% safety score, validating their exceptional protective capabilities in complex risk scenarios. This research provides an effective engineering pathway for building high-security, high-trust LLM applications.

cs.AI↗

arXiv2025

Mock Observations for the CSST Mission: HSTDM--Synthetic Data Generation

The High Sensitivity Terahertz Detection Module (HSTDM), a key component of the backend modules on board the China Space Station Telescope (CSST), will offer great opportunities for the discovery of Terahertz Astronomy, with implications that extend well beyond China to the global astronomical community. It is imperative that the raw data collected by HSTDM undergoes meticulous calibration and processing through the HSTDM data processing pipeline (HSTDM pipeline for short) to ensure the accuracy and effectiveness of the final science data to be archived for further research. This process necessitates that the HSTDM pipeline address instrumental artifacts and effects as well as the coordination of data flow of the scheduled observing sequences under all observing modes of HSTDM within the CSST automated processing environment. As the understanding of CSST HSTDM data processing develops during the pipeline development stage, it becomes essential to assess the accuracy, the robustness and the performance of the HSTDM pipeline under all observing modes of HSTDM so that components of the HSTDM pipeline be rationally added, removed, amended or extended within the modular framework. In this paper, we develop practical simulation methods to facilitate this need. The contribution of synthetic data generation of HSTDM observation includes two parts: 1. HSTDM instrumental effect simulation based on both real testing profiles and simulated models; 2. Observing data flow generation based on HSTDM observing mode scenario. The simulation methods have been implemented and shown to be practical in testing the HSTDM pipeline during the development stage.

astro-ph.IM↗

arXiv2025

Low-Discrepancy Set Post-Processing via Gradient Descent

The construction of low-discrepancy sets, used for uniform sampling and numerical integration, has recently seen great improvements based on optimization and machine learning techniques. However, these methods are computationally expensive, often requiring days of computation or access to GPU clusters. We show that simple gradient descent-based techniques allow for comparable results when starting with a reasonably uniform point set. Not only is this method much more efficient and accessible, but it can be applied as post-processing to any low-discrepancy set generation method for a variety of standard discrepancy measures.

math.OC↗