Searcharxiv⌕ Search

arXiv subjects

Jiayi Huang

Publications and source records attributed to Jiayi Huang.

At least 19 recordsLinked to original sources
arXiv2026

LaSR: Context-Aware Speech Recognition via Latent Reasoning

Speech recognition in specialized domains requires leveraging contextual or topical information to improve the recognition of domain-specific entities. Speech Large Language Models (Speech LLMs) have substantially advanced speech understanding and reasoning capabilities, making context-aware speech recognition possible without predefined bias lists. In this paper, we propose LaSR (Latent Speech Reasoning), a novel training paradigm featuring a context-aware reasoning trajectory that leverages the latent reasoning process. Instead of generating explicit intermediate tokens, LaSR aligns chain-of-thought (CoT) supervision around the acoustic feature region of the target word, and introduces latent reasoning periods for context information grounding and transcriptional transition. Furthermore, to effectively benchmark context-aware speech recognition, we propose Spoken Darwin-Science, a large-scale corpus focusing on academic terminologies. Preliminary experiments on Fun-Audio-Chat demonstrate that LaSR significantly improves terminology recognition without introducing additional latency and consistently outperforms standard supervised fine-tuning baselines. Our findings highlight the potential of latent reasoning in building efficient, context-aware speech assistants.

cs.CL↗
arXiv2026

Phase-Stable Hologram Updates for Large-Scale Neutral-Atom Array Reconfiguration

Dynamic holographic optical tweezers provide a programmable route to array assembly and reconfiguration essential for scalable neutral-atom quantum computation. However, phase mismatch between successive holograms can cause destructive interference during finite spatial light modulator (SLM) refresh. In this work, we analyze finite SLM refresh to establish a phase-stability criterion for prescribed trap amplitudes and develop a weighted-projective Gerchberg-Saxton (WPGS) method that efficiently approximates the corresponding phase-only complex-field optimization. Enforcing this phase constraint also reduces the number of iterations required for each update, achieving convergence within five iterations and enabling hologram generation within a few milliseconds. Numerical simulations of 2D and 3D reconfiguration involving more than $10^3$ traps, together with nonuniform-intensity interlayer transport, show that WPGS preserves endpoint intensity quality while suppressing inter-frame phase mismatch, transient intensity degradation, trap splitting, and motional heating. Similar suppression of phase mismatch and transient-intensity degradation is maintained under Gaussian SLM-plane illumination. These results establish the phase-stability criterion as a practical design principle for dynamic holographic control and scalable neutral-atom array reconfiguration.

quant-ph↗
arXiv2026

$S^3$-Bench: Evaluating Speech Interaction Models as Scientific Voice Assistants

The advance of multimodal large language models (MLLMs) has fundamentally reshaped the paradigm of human-computer interaction, especially speech interaction models capable of seamless conversations. Despite remarkable performance as general voice assistants, their performance in specialized domains remains underexplored, particularly in scientific areas. Scientific interactions introduce formidable challenges, involving rare technical terminology, spoken norms of abbreviations, and the natural verbalization of symbolic special expressions. In this paper, we introduce S$^3$-Bench, a systematic evaluation framework covering 10 major disciplines, consisting of a Knowledge set for speech question-answering and a Dialogue set for multi-turn progressive interactions with simulated user agents. By decomposing a complete atomic turn into stages of speech recognition, perception, knowledge utilization with reasoning, and response pronunciation, we systematically characterize the common challenges and performance tradeoffs of existing approaches. Furthermore, experiments on multi-turn interactions reveal persistent limitations in user adaptation and the generation of accurate, comprehensive, and efficient responses.

cs.CL↗
arXiv2026

NCP-ArchPreview Technical Report: Moving towards Latent Space Language Models through Next Concept Prediction

We introduce NCP-ArchPreview, a latent-space language model that pushes autoregressive pretraining beyond standard next-token prediction (NTP). Alongside NTP, the model learns through Next Concept Prediction (NCP) to predict discrete concepts that span multiple tokens, introducing an explicit and more challenging concept-level objective while preserving standard token-level autoregressive generation. NCP-ArchPreview builds a latent space by constructing a product-quantized concept vocabulary directly from its hidden states, and subsequently learns to predict future concepts via a dedicated Concept Module. These predicted concepts are then fed back to the token level to guide subsequent generation, with NTP and NCP trained jointly end-to-end. We scale this architecture to 8.9B parameters and train it on 5.73T tokens from the Dolma-3 dataset, marking the largest demonstration of a latent-space language model to date. Remarkably, by consuming only 51.3% of the total training tokens, NCP-ArchPreview achieves the final pretraining loss of OLMo-3-7B. Following full pretraining, it outperforms OLMo-3-7B by 2.45 points on the downstream macro-average, including a notable 5.99-point gain on GSM8K. Controlled experiments isolate a clear progression of performance gains stemming from both the latent architecture and the NCP objective. Furthermore, utilizing only 85% of the standard computation, NCP-ArchPreview approaches the training loss of a strictly parameter-aligned 8.9B baseline. The learned latent space remains highly valuable after the pretraining stage: updating just the 17M-parameter VQ module yields a novel, lightweight interface for domain adaptation, while a simple injection of concept representations into a DFlash2 drafter improves the mean accepted length by 4.17% with negligible overhead.

cs.CL↗
arXiv2026

Modulation of charge density waves in a twisted vortex moire superlattice

Twisted moire superlattices in van-der-Waals heterostructures provide a powerful platform for engineering correlated states through moire-band reconstruction. However, whether globally coherent electronic orders can be continuously manipulated at the nanoscale remains largely unexplored. Reconstructed moire structures in small-angle and near-commensurate regime feature continuously varying local environments, offering new opportunities for nanoscale manipulation of correlated phases. Here, we report the modulation of charge density wave (CDW) states in a twisted vortex moire superlattice formed between monolayer VTe2 and superconducting NbSe2. Scanning tunneling microscopy/spectroscopy reveals that the intrinsic long-range CDW of monolayer VTe2 is reconstructed into inequivalent local phases with distinct stability and coherence within a single moire unit cell, including suppressed CDW order and enhanced short-range CDW correlations persisting to room temperature. First-principles calculations show that the reconstructed CDW landscape originates from strong local strain variation, where compressive strain substantially stabilizes the charge order. Furthermore, the modulated CDW states exhibit competing interplay with proximity-induced superconductivity. Our results establish vortex moire superlattices as a versatile platform for nanoscale manipulation of correlated electronic orders in low-dimensional quantum materials.

cond-mat.mtrl-sci↗
arXiv2026

Reliable Inference in Edge-Cloud Model Cascades via Conformal Alignment

Edge intelligence enables low-latency inference via compact on-device models, but assuring reliability remains challenging. We study edge-cloud cascades that must preserve conditional coverage: whenever the edge returns a prediction set, it should contain the true label with a user-specified probability, as if produced by the cloud model. We formalize conditional coverage with respect to the cloud predictive distribution, and introduce a conformal alignment-based (CAb) cascading mechanism that certifies this property with user control over the risk level. Our method casts escalation from edge to cloud models as a multiple-hypothesis testing (MHT) problem, tailoring conformal alignment (CA) to select which inputs can be safely handled at the edge. The proposed CAb model cascading method yields statistical guarantees on the average fraction of edge decisions that satisfy cloud-level conditional coverage. The procedure applies to arbitrary edge prediction sets, including variants of conformal prediction (CP), and exposes a tunable trade-off among coverage, deferral rate, and set size. Experiments on CIFAR-100 image classification and the TeleQnA question-answering (QA) benchmark show that the proposed CAb cascade maintains the target conditional coverage for edge predictions while substantially reducing offloading to the cloud and incurring modest increases in prediction-set size.

cs.LG↗
arXiv2026

PACE: Adaptive Budget Allocation for Time-Efficient Embodied Planning

Reasoning-enhanced large language models have achieved remarkable improvements in planning tasks, yet their deployment in embodied systems remains impractical due to prohibitive inference delays-often exceeding minutes per planning instance. The fundamental bottleneck stems from the serial nature of existing paradigms: models must complete all reasoning before any action execution, leaving execution time windows entirely unexploited. We introduce PACE (Planning with Adaptive Cognitive Effort), a framework that enables interleaved reasoning and execution through two key innovations: an Interleaved Think-Act architecture that pipelines cognitive processing with action execution, and a Dynamic Budget Allocator that adapts reasoning token budgets to available execution time windows. On the Robotouille benchmark using Qwen3-8B-AWQ, PACE achieves a 10% success rate-representing a 67% improvement over the ReAct+Think baseline-while delivering 6.9 times acceleration in thinking time compared to unconstrained reasoning. The framework hides 66.8% of thinking time within execution windows, demonstrating that strategic cognitive effort allocation can simultaneously improve both planning quality and time efficiency. These results provide evidence that time-aware architectural innovations enable reasoning models to operate in latency-sensitive embodied domains where they were previously impractical.

cs.RO↗
arXiv2026

Distributionally robust two-stage model predictive control: adaptive constraint tightening with stability guarantee

This paper proposes a two-stage distributionally robust model predictive control (TSDR-MPC) scheme for stochastic disturbances with unknown time-varying means and covariances. By defining a Wasserstein ambiguity set on the disturbance-to-constraint space, constraint violation penalties are formulated as a second-stage problem, enabling adaptive tightening. A finitely convergent cutting-plane algorithm is developed for real-time implementation. The framework naturally degrades to deterministic MPC as uncertainty vanishes, without pre-specified tightening parameters. Theoretical guarantees include feasibility, finite-time termination, and an asymptotic average cost bound. Numerical simulations validate its adaptability and robustness.

eess.SY↗
arXiv2026

MotionPRO: Exploring the Role of Pressure in Human MoCap and Beyond

Existing human Motion Capture (MoCap) methods mostly focus on the visual similarity while neglecting the physical plausibility. As a result, downstream tasks such as driving virtual human in 3D scene or humanoid robots in real world suffer from issues such as timing drift and jitter, spatial problems like sliding and penetration, and poor global trajectory accuracy. In this paper, we revisit human MoCap from the perspective of interaction between human body and physical world by exploring the role of pressure. Firstly, we construct a large-scale human Motion capture dataset with Pressure, RGB and Optical sensors (named MotionPRO), which comprises 70 volunteers performing 400 types of motion, encompassing a total of 12.4M pose frames. Secondly, we examine both the necessity and effectiveness of the pressure signal through two challenging tasks: (1) pose and trajectory estimation based solely on pressure: We propose a network that incorporates a small kernel decoder and a long-short-term attention module, and proof that pressure could provide accurate global trajectory and plausible lower body pose. (2) pose and trajectory estimation by fusing pressure and RGB: We impose constraints on orthographic similarity along the camera axis and whole-body contact along the vertical axis to enhance the cross-attention strategy to fuse pressure and RGB feature maps. Experiments demonstrate that fusing pressure with RGB features not only significantly improves performance in terms of objective metrics, but also plausibly drives virtual humans (SMPL) in 3D scene. Furthermore, we demonstrate that incorporating physical perception enables humanoid robots to perform more precise and stable actions, which is highly beneficial for the development of embodied artificial intelligence. Project page is available at: https://nju-cite-mocaphumanoid.github.io/MotionPRO/

cs.CV↗
arXiv2026

Capacity-Aware Inference: Mitigating the Straggler Effect in Mixture of Experts

The Mixture of Experts (MoE) is an effective architecture for scaling large language models by leveraging sparse expert activation to balance performance and efficiency. However, under expert parallelism, MoE suffers from inference inefficiencies due to imbalanced token-to-expert assignment, where underloaded experts complete computations early but must wait for overloaded experts, leading to global delays. We define this phenomenon as the \textbf{\textit{Straggler Effect}}, as the most burdened experts dictate the overall inference latency. To address this, we first propose \textit{\textbf{Capacity-Aware Token Drop}}, which enforces expert capacity limits by discarding excess tokens from overloaded experts, effectively reducing load imbalance with minimal performance impact (e.g., $30\%$ speedup with only $0.9\%$ degradation on OLMoE). Next, given the presence of low-load experts remaining well below the capacity threshold, we introduce \textit{\textbf{Capacity-Aware Expanded Drop}}, which allows tokens to include additional local experts in their candidate set before enforcing strict local capacity constraints, thereby improving load balance and enhancing the utilization of underused experts. Extensive experiments on both language and multimodal MoE models demonstrate the effectiveness of our approach, yielding substantial gains in expert utilization, model performance, and inference efficiency, e.g., applying Expanded Drop to Mixtral-8$\times$7B-Instruct yields a {0.2\%} average performance improvement and a {1.85$\times$} inference speedup. The code is released at: https://github.com/CASE-Lab-UMD/Capacity-Aware-MoE.

cs.LG↗
arXiv2026

Foundations of Reliable Inference: Reliability-Efficiency Co-Design

Reliable inference requires that artificial intelligence (AI) models provide trustworthy uncertainty estimates, not merely accurate predictions. Recent advances in Bayesian learning have made significant progress toward this goal, and growing concerns about computational overhead have jointly shifted the design criterion from reliability alone to the co-design of reliability and efficiency, i.e., reducing computational overhead while preserving trustworthy uncertainty quantification. This thesis develops a unified framework from two perspectives to address the central question: can we efficiently perform reliable inference?

cs.LG↗
arXiv2026

ReaLB: Real-Time Load Balancing for Multimodal MoE Inference

Mixture-of-Experts (MoE) architectures are widely used in modern large language models and multimodal models. However, inference efficiency is often limited by highly dynamic and skewed expert workloads across different modalities. During the prefill stage with large batch sizes, vision tokens frequently dominate the input sequences. Under expert parallelism (EP), this leads to severe load imbalance, where a subset of devices becomes overloaded, reducing overall system throughput. We propose ReaLB, a real-time load balancing method for multimodal MoE (MMoE) inference that introduces zero scheduling overhead. ReaLB dynamically adjusts the computation precision of MoE experts at runtime on a per-EP-rank basis. For ranks dominated by vision-heavy experts, ReaLB assigns lower-precision computation to improve execution efficiency by exploiting FP4 Tensor Cores. ReaLB does not require redundant experts or additional memory allocation. Instead, it performs layer-wise expert precision transformation on the fly and hides the associated overhead within the dispatch phase before MoE computation. Experiments on representative MMoE models show that ReaLB achieves 1.10$\times$-1.32$\times$ end-to-end speedup while limiting average accuracy degradation to within 1%.

cs.DC↗
arXiv2026

Improved third-order scheme in pseudopotential lattice Boltzmann model for multiphase flows

The lattice Boltzmann (LB) equation with a third-order scheme can be regarded as a unified and self-consistent framework of the pseudopotential LB model for multiphase flows. In this work, we theoretically analyze pseudopotential LB simulations of two-phase Poiseuille flow at the discrete level. The finite-difference velocity equation is derived for both grid-aligned and grid-oblique cases. The terms responsible for spurious velocity oscillations near the phase interface are identified. Based on this discrete-level analysis, an improved third-order scheme is proposed to suppress spurious velocity oscillations. This scheme does not introduce any additional conceptual or computational complexity compared with the original one and reduces to the original scheme under static conditions. Numerical simulations of two-phase Poiseuille flow validate the present theoretical analysis and demonstrate the effectiveness of the improved scheme. Then, annular shear flow with a curved phase interface is considered to show that spurious velocity oscillations can also be effectively suppressed by the improved scheme in cases with such interfaces. Finally, the falling of a droplet in a vertical channel is simulated, and the results show that spurious velocity oscillations can lead to an overestimation of the drag force and distinct falling patterns. These results highlight the necessity of using the improved third-order scheme to suppress spurious oscillations and obtain reliable results.

physics.flu-dyn↗
arXiv2026

Optimized Certainty Equivalent Risk-Controlling Prediction Sets

In safety-critical applications such as medical image segmentation, prediction systems must provide reliability guarantees that extend beyond conventional expected loss control. While risk-controlling prediction sets (RCPS) offer probabilistic guarantees on the expected risk, they fail to capture tail behavior and worst-case scenarios that are crucial in high-stakes settings. This paper introduces optimized certainty equivalent RCPS (OCE-RCPS), a novel framework that provides high-probability guarantees on general optimized certainty equivalent (OCE) risk measures, including conditional value-at-risk (CVaR) and entropic risk. OCE-RCPS leverages upper confidence bounds to identify prediction set parameters that satisfy user-specified risk tolerance levels with provable reliability. We establish theoretical guarantees showing that OCE-RCPS satisfies the desired probabilistic constraint for loss functions such as miscoverage and false negative rate. Experiments on image segmentation demonstrate that OCE-RCPS consistently meets target satisfaction rates across various risk measures and reliability configurations, while OCE-CRC fails to provide probabilistic guarantees.

cs.LG↗
arXiv2026

Hemlet: A Heterogeneous Compute-in-Memory Chiplet Architecture for Vision Transformers with Group-Level Parallelism

Vision Transformers (ViTs) have established new performance benchmarks in vision tasks such as image recognition and object detection. However, these advancements come with significant demands for memory and computational resources, presenting challenges for hardware deployment. Heterogeneous compute-in-memory (CIM) accelerators have emerged as a promising solution for enabling energy-efficient deployment of ViTs. Despite this potential, monolithic CIM-based designs face scalability issues due to the size limitations of a single chip. To address this challenge, emerging chiplet-based techniques offer a more scalable alternative. However, chiplet designs come with their own costs, as they introduce expensive communication, which can hinder improvements in throughput. This work introduces Hemlet, a heterogeneous CIM chiplet system designed to accelerate ViT workloads. Hemlet enables flexible resource scaling through the integration of heterogeneous analog CIM (ACIM), digital CIM (DCIM), and Intermediate Data Process (IDP) chiplets. To improve throughput while reducing communication overhead, it employs a group-level parallelism (GLP) mapping strategy and system-level dataflow optimization, achieving speedups ranging from 2.41x to 5.74x across various hardware configurations within the chiplet system. Our evaluation results show that Hemlet can reach a throughput of 9.56 TOPS with an energy efficiency of 4.98 TOPS/W.

cs.AR↗
arXiv2026

A primer on treatment planning aspects for temporally modulated pulsed radiation therapy

Temporally modulated pulsed radiotherapy (TMPRT) delivers conventional fraction doses of radiation using temporally separated pulses of low doses (<30 cGy) yielding fraction-effective dose rates of around 6.7 cGy/min with the goal to exploit tumor radiation hypersensitivity, which was observed in both, preclinical models and in human clinical trials. To facilitate TMPRT, volumetric modulated arc therapy (VMAT) and 3D-CRT planning techniques were developed following the guidelines of the proposed NRG CC-017 trial. Plans were evaluated with respect to homogeneity, conformality, and adherence to dose constraints. Deliverability of plans was assessed using in-phantom measurements for absorbed dose accuracy at low dose rates and using EPID for isodose verification. For VMAT only single arc plans were found to be acceptable due to otherwise unacceptably heterogeneous field doses, while for dynamic conformal arcs machine limtations on the number of monitor units per degree require the use of partial arcs for each pulse. Delivery of plans at low dose rates (< 100 MU/min) was accurate with high Gamma pass rates on modern LINACs and moderate pass rates on legacy LINACs, in line with their general performance. Generally, VMAT is preferred to achieve optimal homogeneity, conformality, and organ-at-risk sparing, while the use of 3D-CRT can increase the availability of TMPRT for more patients and clinics.

physics.med-ph↗
arXiv2025

Calibrating Bayesian Learning via Regularization, Confidence Minimization, and Selective Inference

The application of artificial intelligence (AI) models in fields such as engineering is limited by the known difficulty of quantifying the reliability of an AI's decision. A well-calibrated AI model must correctly report its accuracy on in-distribution (ID) inputs, while also enabling the detection of out-of-distribution (OOD) inputs. A conventional approach to improve calibration is the application of Bayesian ensembling. However, owing to computational limitations and model misspecification, practical ensembling strategies do not necessarily enhance calibration. This paper proposes an extension of variational inference (VI)-based Bayesian learning that integrates calibration regularization for improved ID performance, confidence minimization for OOD detection, and selective calibration to ensure a synergistic use of calibration regularization and confidence minimization. The scheme is constructed successively by first introducing calibration-regularized Bayesian learning (CBNN), then incorporating out-of-distribution confidence minimization (OCM) to yield CBNN-OCM, and finally integrating also selective calibration to produce selective CBNN-OCM (SCBNN-OCM). Selective calibration rejects inputs for which the calibration performance is expected to be insufficient. Numerical results illustrate the trade-offs between ID accuracy, ID calibration, and OOD calibration attained by both frequentist and Bayesian learning methods. Among the main conclusions, SCBNN-OCM is seen to achieve best ID and OOD performance as compared to existing state-of-the-art approaches at the cost of rejecting a sufficiently large number of inputs.

cs.LG↗
arXiv2025

Distilling Calibration via Conformalized Credal Inference

Deploying artificial intelligence (AI) models on edge devices involves a delicate balance between meeting stringent complexity constraints, such as limited memory and energy resources, and ensuring reliable performance in sensitive decision-making tasks. One way to enhance reliability is through uncertainty quantification via Bayesian inference. This approach, however, typically necessitates maintaining and running multiple models in an ensemble, which may exceed the computational limits of edge devices. This paper introduces a low-complexity methodology to address this challenge by distilling calibration information from a more complex model. In an offline phase, predictive probabilities generated by a high-complexity cloud-based model are leveraged to determine a threshold based on the typical divergence between the cloud and edge models. At run time, this threshold is used to construct credal sets -- ranges of predictive probabilities that are guaranteed, with a user-selected confidence level, to include the predictions of the cloud model. The credal sets are obtained through thresholding of a divergence measure in the simplex of predictive probabilities. Experiments on visual and language tasks demonstrate that the proposed approach, termed Conformalized Distillation for Credal Inference (CD-CI), significantly improves calibration performance compared to low-complexity Bayesian methods, such as Laplace approximation, making it a practical and efficient solution for edge AI deployments.

cs.LG↗