arXiv subjects

Hao Huang

Publications and source records attributed to Hao Huang.

At least 19 recordsLinked to original sources

arXiv2026

FastMap: Real-Time Semantic Map Completion via Bitwise Masked Modeling

Semantic map completion, which predicts the layout of unobserved regions from partial observations, is a critical capability for indoor robot navigation. Existing approaches either rely on high-dimensional discrete codebooks that inflate memory, or on iterative diffusion sampling that is too slow for real-time use. We present FastMap, a lightweight two-stage framework for completing top-down categorical semantic maps. First, a lookup-free BitVAE exploits the inherently binary (one-hot) structure of semantic maps to compress each map patch into compact bitwise tokens, yielding a 0.41GB model that is 3.7 times smaller than the prior masked-modeling baseline. Second, a Masked AutoEncoder (MAE)-style transformer reconstructs missing tokens in a single forward pass at 0.011s/map. To support object goal navigation, we additionally introduce an object-aware masking strategy that masks the target category during training and conditions generation on a learnable category embedding, without adding inference cost. On the Gibson benchmark, FastMap achieves 34.10% mIoU and 45.84% semantic Success Rate (sSR), more than doubling the previous best (21.88%), and reaches 83.8% Success Rate on downstream ObjectNav, the highest among the compared navigation methods.

cs.RO↗

arXiv2026

CTC-TTS: LLM-based dual-streaming text-to-speech with CTC alignment

Large-language-model (LLM)-based text-to-speech (TTS) systems can generate natural speech, but most are not designed for low-latency dual-streaming synthesis. High-quality dual-streaming TTS depends on accurate text--speech alignment and well-designed training sequences that balance synthesis quality and latency. Prior work often relies on GMM-HMM based forced-alignment toolkits (e.g., MFA), which are pipeline-heavy and less flexible than neural aligners; fixed-ratio interleaving of text and speech tokens struggles to capture text--speech alignment regularities. We propose CTC-TTS, which replaces MFA with a CTC based aligner and introduces a bi-word based interleaving strategy. Two variants are designed: CTC-TTS-L (token concatenation along the sequence length) for higher quality and CTC-TTS-F (embedding stacking along the feature dimension) for lower latency. Experiments show that CTC-TTS outperforms fixed-ratio interleaving and MFA-based baselines on streaming synthesis and zero-shot tasks.

eess.AS↗

arXiv2026

On cancellative pairs of families of subsets

A pair $(\mathcal{A}, \mathcal{B})$ of families of subsets of $[n]$ is cancellative if whenever $A, A' \in \mathcal{A}, B \in \mathcal{B}$ satisfy $A \cup B=A' \cup B$, then $A=A'$, and whenever $A \in \mathcal{A}, B, B' \in \mathcal{B}$ satisfy $A \cup B=A \cup B'$, then $B=B'$. We show that for every cancellative pair $(\mathcal{A}, \mathcal{B})$, the inequality $|\mathcal{A}||\mathcal{B}| \le 2.25^n$ holds, matching Tolhuizen's $(2.25-o(1))^n$ lower bound construction.

math.CO↗

arXiv2026

Why Are LLM Backdoor Defenses Fragmented? A Feature-Level Explanation with Sparse Autoencoders

Backdoor attacks pose a serious threat to large language models (LLMs), but existing defenses remain fragmented, failing to pro?vide unified defense against both dirty-label and clean-label attacks. To investigate why such fragmentation arises, we present the first systematic feature-level mechanistic analysis of LLM backdoors using sparse autoencoders (SAEs). Starting from a 2 x 2 comparison of clean and poisoned models on clean and triggered inputs, we trace backdoor-induced logit shifts to high-contributing SAE features and categorize them into four roles: interac?tion, suppressed, mixed, and weight-modified features. This taxonomy reveals system?atic encoding differences: dirty-label back?doors are dominated by isolated interaction features, whereas clean-label backdoors rely more on heterogeneous mixtures of mixed and weight-modified features. These differ?ences explain why existing defenses remain fragmented across attack paradigms. We val?idate this hypothesis through inference-time feature clamping, which reduces ASR to at most 10.8% in most dirty-label settings and at most 15.4% in the majority of clean-label settings, while preserving benign-task perfor?mance. These results show that SAE-based analysis can explain defense fragmentation and guide interpretable backdoor mitigation.

cs.CR↗

arXiv2026

Endpoint Energy Atoms Force Local Pressure Concentration in Three-Dimensional Navier-Stokes Flow

We prove that a point atom in an endpoint kinetic-energy measure of a three-dimensional incompressible Navier-Stokes flow on the flat torus forces quantitative concentration of the actual pressure. At each restart time tau, a same-state constraint-response comparison evolves u(tau) by the pressure-free componentwise advection-diffusion equation driven by u, and subtracts the resulting passive field z_tau. Writing r_tau = u - z_tau and g_tau = Q z_tau, we obtain an exact relative pressure-work identity. If the atom has mass m, Nash smoothing makes z_tau terminally non-atomic while r_tau retains terminal atomic mass at least m; hence, after every fixed restart, the terminal limsup of the accumulated pressure work is at least m/2. Mesoscopic localization yields scale-explicit lower bounds for gauge-invariant L2 pressure oscillation and for the local L2 mass of the pressure gradient in shrinking terminal cylinders. Consequently, on every terminal neighborhood of the atomic point, the pressure modulo functions of time and its gradient fail to be square integrable. The relative pressure-work principle extends to constrained solenoidal Oseen evolutions: as tau approaches T*, the zero-initial responses vanish in weak-tail topologies, although every fixed restart retains the same lower bound. This identifies a necessary pressure-work mechanism, generated by the incompressibility constraint, that links endpoint atoms to local pressure concentration, without requiring a full regularity hypothesis.

math.AP↗

arXiv2026

Focus Where It Counts: A Salience-Driven Vision-Language Model for Low Vision Assistance

Vision-language models (VLMs) are rapidly progressing and offer promising capabilities for assistive technologies supporting persons with blindness or low vision. However, existing VLMs are primarily designed for general-purpose captioning and do not explicitly model human perceptual priorities, thereby limiting their ability to emphasize the most relevant information in a scene. To address this gap, we propose a salience-driven captioning framework that prioritizes scene elements according to their importance for human-centered assistance. We curate three salience-aware datasets, namely, Salience COCO, Salience Flickr, and Salience VizWiz, with object-level salience annotations designed to reflect the visual information most relevant to low vision users across different environments. Building on these datasets, we introduce Salience-LLaVA, a salience-aware VLM that incorporates salience cues to generate captions in which important elements are mentioned in the order of importance. Our work makes four main contributions. We build salience-aware datasets verified by low vision participants, propose Salience-LLaVA to describe objects in the order of importance, introduce SCMI to evaluate ordering accuracy, and deploy the system on assistive glasses to demonstrate real-world practicality. Code and datasets are available at: https://github.com/topo-focus/Topofocus

cs.CV↗

arXiv2026

MeshPriorDiT: Hierarchical Modeling for Action-Conditioned Cloth Dynamics

Action-conditioned cloth dynamics prediction requires both locally plausible deformation and long-range coordination. Existing approaches largely follow two paradigms. Mesh-based GNNs capture local physical responses through material connectivity. However, their finite message-passing range limits coordination between topologically distant regions, while autoregressive rollouts tend to accumulate prediction errors. Transformer-based dynamics models capture long-range interactions through global attention, but often operate without explicit material connectivity and must learn local topological responses directly from data. We propose MeshPriorDiT, a hierarchical dynamics model that decomposes future cloth motion into a structured mesh prior and a generative residual. An action-conditioned mesh GNN first predicts multi-step vertex displacements, yielding a reference trajectory that respects material topology and grasp constraints. Conditioned on historical states, planned actions, and the mesh prior, a Residual DiT then uses conditional flow matching to jointly generate the residual motion not captured by the prior. The generated residual is further rescaled and decoded using material adjacency to coordinate corrections across neighboring vertices. We evaluate MeshPriorDiT on 15-step autoregressive rollouts across three cloth manipulation tasks. Averaged over the three tasks, MeshPriorDiT reduces average Global MSE by 43.42% relative to the GNN-Only baseline and by 75.03% relative to the DiT-DDPM baseline, while maintaining a favorable Edge-strain MSE comparable to that of GNN-Only.

cs.RO↗

arXiv2026

Latent Softmax for Data-Efficient Phoneme-Based Multilingual ASR Across Tonal and Non-Tonal Languages

Phoneme-based multilingual automatic speech recognition (ASR) can share acoustic evidence across languages more directly than language-specific subword modeling. When tonal and non-tonal languages are jointly trained, however, their supervision granularity does not match: tonal languages annotate tone-marked vowels, whereas non-tonal languages typically provide only base-vowel labels. A standard softmax either treats the two as unrelated classes, weakening cross-lingual sharing, or collapses tones, losing distinctions required by tonal languages. We propose Latent Softmax, a connectionist temporal classification (CTC)-compatible output layer that models tone-marked vowels as subclasses and base vowels as major classes, while consonants and the CTC blank remain singleton labels. When only a base-vowel major-class label is observed, the tone-marked vowel subclass is treated as latent and marginalized out. Multilingual experiments on AISHELL-1 Mandarin and LibriSpeech English show that Latent Softmax reduces speech-to-phoneme (S2P) phoneme error rates over a standard softmax multilingual baseline by 8.4% on AISHELL-1, 17.5% on LibriSpeech test-clean, and 12.6% on test-other. The improved speech-to-phoneme encoders also yield consistent word error rate gains for both large-language-model phoneme-to-grapheme conversion (LLM-P2G) and projector-based interfaces. After code-switching adaptation in the evaluated Mandarin--English setting, Latent Softmax reduces projector-based mixed error rate by 2.6% on ASRU2019 and 9.5% on CS-Dialogue, whereas the LLM-P2G results do not establish a consistent advantage.

eess.AS↗

arXiv2026

PdrQC: Pauli-space Discriminative Representations based Quantum Classifier

Quantum classification faces two key challenges. First, the difficulty of distinguishing between different classes varies: some class pairs are easy to separate, while others are more challenging. Second, practical execution is affected by noise, finite sampling, and measurement overhead. To address these issues, we propose the Pauli-Space Discriminative-Representation based Quantum Classifier (PdrQC), a framework for task-adaptive multiclass quantum classification. The method evaluates candidate upload circuits using low-weight Pauli features and formulates upload design as a structured model selection problem based on discriminative representations. By progressively selecting upload structures and compact Pauli readout features for the target multiclass task, the framework achieves a better balance between classification accuracy and resource efficiency. Numerical simulations were conducted on the MNIST and Fashion-MNIST datasets with $K\in\{2,3,5,7,10\}$. The results demonstrate that PdrQC, through its task-adaptive Pauli representation, achieves an effective balance among multiclass classification accuracy, quantum-circuit complexity, and measurement overhead, making it suitable for multiclass quantum classification under limited hardware resources.

quant-ph↗

arXiv2026

Maximizing directed cycles in tournaments

Determining the combinatorial structures that maximize the number of prescribed substructures is a central theme in extremal combinatorics. Grzesik, Král', Lovász and Volec showed that when $\ell$ is not divisible by $4$, the random tournament contains asymptotically the most directed cycles of length $\ell$ among all $n$-vertex tournaments. In the paper, we resolve the remaining cases where $\ell$ is divisible by $4$. We show that, in this regime, the so-called carousel tournament asymptotically maximizes the number of directed $\ell$-cycles among all $n$-vertex tournaments, and in particular contains strictly more such cycles than the random tournament. This confirms the conjecture of Bartley and Day.

math.CO↗

arXiv2026

Full-Tail Dynamical Rigidity Forced by Atomic Navier-Stokes Energy Concentration

Let a smooth, unforced, three-dimensional Navier--Stokes flow on the flat torus approach a finite terminal time. Its kinetic-energy densities converge, along the full time variable, to a unique endpoint measure. We prove that each point atom forces a same-parent, full-tail dynamical rigidity. A preassigned level-crossing catalogue and nested local Hodge projections produce orthogonal packets. From the entire packet tail, we extract a single backward adjoint whose terminal energy concentrates at the atom. Cauchy saturation then locks this adjoint to every sufficiently late packet and yields uniform two-parameter saturation of both the constrained Oseen propagator and its adjoint, together with vanishing first-order dissipation. Consequently, every sufficiently late fixed-root descendant has infinite delayed second-order action and non-integrable positive enstrophy production. Equivalently, a delayed second-order operator budget determined solely by the Navier--Stokes parent must fail arbitrarily close to the endpoint.

math.AP↗

arXiv2026

Exact Lagrangian Realization and Robust Strain Sensing in Incompressible Flow

We prove that the full group $\SL(3,\R)$ occurs as the set of one-particle deformation gradients of periodic, unforced, single-shell solutions of both the three-dimensional Euler and Navier--Stokes equations. More precisely, given a particle label $p\in\T^3$, a time $T>0$, and $F_*\in\SL(3,\R)$, every sufficiently large odd integer $N$ admits a real-analytic curl eigenfield $W_N$ with $\operatorname{curl}W_N=NW_N$. With an explicit scalar amplitude, $W_N$ yields a steady Euler solution; with an explicit exponentially decaying amplitude, it yields a Navier--Stokes solution for any positive viscosity, and in both cases $\nabla_aX(p,T)=F_*$. The lifted particle trajectory is an embedded analytic arc with nowhere-vanishing velocity. The construction combines global trace-free symmetric-matrix control on $\SL(3,\R)$, a Beltrami Cauchy problem along the controlled arc, Runge approximation, inverse localization on the torus, and a finite-dimensional endpoint correction. We also classify finite material-direction systems that determine every trace-free strain after arbitrary volume-preserving deformation. In dimension $n$, this congruence-robust property holds exactly when the associated rank-one projectors span $\Sym(n)$; hence the sharp number of scalar channels is $n(n+1)/2$. In dimension three, among minimal systems the undeformed outer-product lower bound is maximized exactly by the six axes of a regular icosahedron. The realization theorem shows that the full deformation-group quantifier in this sensing result is dynamically attained within the rigid class above.

math.AP↗

arXiv2026

Glass Surface Detection: Leveraging Reflection Dynamics in Flash/No-flash Imagery

Glass surfaces are ubiquitous in daily life, typically appearing colorless, transparent, and lacking distinctive features. These characteristics make glass surface detection a challenging computer vision task. Existing glass surface detection methods always rely on boundary cues (\textit{e.g.}, window and door frames) or reflection cues to locate glass surfaces, but they fail to fully exploit the intrinsic properties of the glass itself for accurate localization. We observed that in most real-world scenes, the illumination intensity in front of the glass surface differs from that behind it, which results in variations in the reflections visible on the glass surface. Specifically, when standing on the brighter side of the glass and applying a flash towards the darker side, existing reflections on the glass surface tend to disappear. Conversely, while standing on the darker side and applying a flash towards the brighter side, distinct reflections will appear on the glass surface. Based on this phenomenon, we propose \textit{NFGlassNet}, a novel method for glass surface detection that leverages the reflection dynamics present in flash/no-flash imagery. Specifically, we propose a Reflection Contrast Mining Module (RCMM) for extracting reflections, and a Reflection Guided Attention Module (RGAM) for fusing features from reflection and glass surface for accurate glass surface detection. For learning our network, we also construct a dataset consisting of $\sim$3.3$K$ no-flash and flash image pairs captured from various scenes with corresponding ground truth annotations. Extensive experiments demonstrate that our method outperforms the state-of-the-art methods. Our code, model, and dataset will be available upon acceptance of the manuscript.

cs.CV↗

arXiv2026

MVGD-Net: A Novel Motion-aware Video Glass Surface Detection Method

Glass surface ubiquitous in both daily life and professional environments presents a potential threat to vision-based systems, such as robot and drone navigation. To solve this challenge, most recent studies have shown significant interest in Video Glass Surface Detection (VGSD). We observe that objects in the reflection (or transmission) layer appear farther from the glass surfaces. Consequently, in video motion scenarios, the notable reflected (or transmitted) objects on the glass surface move slower than objects in non-glass regions within the same spatial plane, and this motion inconsistency can effectively reveal the presence of glass surfaces. Based on this observation, we propose a novel network, named MVGD-Net, for detecting glass surfaces in videos by leveraging motion inconsistency cues. Our MVGD-Net features three novel modules: the Cross-scale Multimodal Fusion Module (CMFM) that integrates extracted spatial features and estimated optical flow maps, the History Guided Attention Module (HGAM) and Temporal Cross Attention Module (TCAM), both of which further enhances temporal features. A Temporal-Spatial Decoder (TSD) is also introduced to fuse the spatial and temporal features for generating the glass region mask. Furthermore, for learning our network, we also propose a large-scale dataset, which comprises 312 diverse glass scenarios with a total of 19,268 frames. Extensive experiments demonstrate that our MVGD-Net outperforms relevant state-of-the-art methods.

cs.CV↗

arXiv2026

Muon as a Residual Connection

Muon has recently emerged as one of the most effective optimizers for training large neural networks, yet its empirical success has been explained from several different perspectives. In this paper, we propose a simple mechanistic interpretation: Muon can be understood as an implicit residual connection during training. Specifically, orthogonalizing the update can sacrifice some immediate gradient fidelity while improving representation preservation for downstream layers. We study this trade-off in controlled linear optimization settings, where Muon can learn representations that are slower to fit a local target but easier for downstream layers to exploit. Our results suggest a conceptual explanation for Muon and a design perspective for optimizers that balance local descent with downstream usability.

cs.LG↗

arXiv2026

High-accuracy Low-Bit KV-Cache Quantization via Local Distribution Restoration

Long-context large language model inference relies on the KV cache to avoid redundant attention computation, but incurs high memory and bandwidth overheads. Low-bit KV-cache quantization reduces this cost, yet it severely degrade quality; particularly, one-bit quantization reduces accuracy from 84.2% to 47.8% on Llama-3.1-8B under RULER. Rather than common beliefs that absolute error of logits, we find that the root cause is structured local misranking, where the distribution of logits in top-K region is drifted. We thereby propose local distribution restoration, a new technique that detects steps with high local distribution risk from quantized-logit features and restores only the selected top-K candidate distribution before token selection. We implement DGAP to achieve local distribution restoration, with efficient risk detcetors and correctors. Expeirments show that on Llama-3.1-8B, DGAP recovers K1V1 RULER accuracy from 47.8% to 83.2% and reduces distribution drift from 0.38 to 0.14; across Llama, Mistral, and Qwen models, it preserves the persistent low-bit KV-cache footprint with modest decode overhead.

cs.LG↗

arXiv2026

DualEval: Joint Model-Item Calibration for Unified LLM Evaluation

Current LLM evaluation relies on two complementary but often disconnected signals: static benchmarks with objective correctness labels and arena-style preference data that better reflect open-ended user interactions. We introduce DualEval, a latent model-item calibration framework that represents models and evaluation items in a shared space, jointly estimating model ability together with item difficulty and sharpness. We apply DualEval across four domains: coding, math, miscellaneous domain-knowledge tasks, and generic everyday user queries. Our evaluation uses 18 frontier LLMs, static benchmark labels, and reward-model scores validated against held-out human preferences for open-ended model responses. Empirically, our framework produces reliable and balanced model rankings, and its learned item-level profiles support downstream applications such as benchmark compression for sample-efficient evaluation and anomaly detection for contamination or outlier analysis. Overall, DualEval unifies static and arena-style evaluation through joint model-item calibration, producing model rankings and item-level diagnostics that support more sample-efficient, interpretable, and auditable evaluation pipelines.

cs.LG↗

arXiv2026

SIMBA: ABidirectional Retrieval Forward Simulation Framework for Modeling FY-4A GIIRS Hyperspectral Infrared Radiances Toward NWP Applications

Hyperspectral infrared observations are an important data source for numerical weather prediction (NWP) because they provide rich information on the vertical structure of atmospheric temperature and humidity. However, most existing deep learning methods mainly focus on one-way retrieval from radiances to atmospheric profiles, while the reverse radiance simulation process and the consistency between atmospheric state space and radiance observation space are insufficiently considered. In this study, we propose SIMBA, a unified bidirectional retrieval-forward simulation framework for FY-4A GIIRS hyperspectral infrared radiance modeling toward NWP applications. The framework jointly performs atmospheric profile retrieval and radiance reconstruction, introduces a cycle-consistency constraint to strengthen the coupling between the two processes, and employs a bidirectional Mamba state-space module to capture long-range dependencies along pressure levels. Using collocated FY-4A GIIRS observations and ERA5 reanalysis data, the proposed method is evaluated for temperature retrieval, specific humidity retrieval, long-wave radiance reconstruction, and medium-wave radiance reconstruction. Experimental results show that SIMBA outperforms several representative deep learning baselines across both retrieval and reconstruction tasks, while ablation experiments confirm the contribution of the bidirectional design and cycle-consistency mechanism. These results demonstrate that the proposed framework is effective for joint atmospheric profile retrieval and hyperspectral infrared radiance modeling, and suggest potential for future Jacobian-related analysis and NWP-oriented extensions.

eess.IV↗