Searcharxiv⌕ Search

arXiv subjects

Tianhao Wu

Publications and source records attributed to Tianhao Wu.

At least 19 recordsLinked to original sources
arXiv2026

Tan's Contact for Trapped Spin-Orbit-Coupled Fermi Gases at Finite Temperature

Tan's contact, bridging microscopic features of many-body quantum systems to two-body contact interactions, is pivotal in the description of quantum gases. However, past studies of Tan's contact of the spin-orbit coupled Fermi gases have been exclusively discussed at zero temperature. The thermal effects on that, which cannot be neglected in most experimental conditions, has not been explored so far. Here, through a path integral Monte Carlo (PIMC) study, we show that thermal effects fundamentally alter the scaling relation between spin-orbit coupling (SOC) and Tan's contact. The linear scaling of Tan's contact with respect to the strength of SOC found at zero temperature is drastically undermined by the thermal effects. And a new mechanism based on the competition between the mean occupancy and the energy-reweighted occupancy of the opposite spin fermion pairs within the interaction range has been unveiled, which radically changes the above linear scaling to a non-monotonic dependence. Our work should be useful for future experiments to explore the preternatural thermal effects on Tan's contact of the strongly interacting spin-orbit-coupled Fermi gas.

cond-mat.quant-gas↗
arXiv2026

SeRV: Semantic-Aligned Residual Vector Quantization for American Sign Language Generation

American Sign Language (ASL) generation remains challenging due to limited paired text-ASL motion data and the difficulty of learning motion representations both precise for reconstruction and predictable from linguistic input. Existing methods rely on motion tokenizers optimized for reconstruction, without explicit semantic supervision from paired text. As a result, the learned tokens remain limited in supporting semantically consistent and fine-grained ASL motion generation. To address this limitation, we propose SeRV (Semantic-Aligned Residual Vector Quantization), a semantic-aligned RVQ tokenizer for ASL generation. SeRV learns a semantically structured residual token space by combining global-level motion-text alignment with token-level text-conditioned supervision. Building on this tokenizer, a Hierarchical GPT predicts residual motion tokens in a coarse-to-fine manner, generating structurally coherent and semantically aligned 3D ASL motion. We further construct a large-scale reconstructed 3D ASL motion-text benchmark by recovering paired 3D motion from OpenASL videos. Experiments on How2Sign and OpenASL show that SeRV outperforms six state-of-the-art baselines, with user studies assessing reconstruction quality and semantic fidelity. Samples of our reconstructed 3D ASL motion are available on our project page: https://signvideos.github.io/Demo/.

cs.CV↗
arXiv2026

EMG-FM-Bench: A Comprehensive Benchmark for Foundation Model Transfer and Adaptation on Electromyography

Foundation models (FMs) are increasingly being developed for general time series and physiological signals, yet their transferability to downstream physiological tasks remains poorly understood. This question is particularly challenging for electromyography (EMG), where signal distributions vary substantially across users, sensing configurations, acquisition hardware, and downstream tasks. We introduce EMG-FM-Bench, a systematic benchmark for studying foundation-model transfer and adaptation on EMG. EMG-FM-Bench unifies 20 public datasets with over 1 million EMG segments and evaluates nine pretrained foundation models across four questions: how pretrained models perform when frozen or fully fine-tuned, how much pretraining helps compared with training the same model from scratch, how well models generalize to new users with limited labeled data, and how performance changes across different EMG tasks. Across the benchmark, linear probing provides useful information about pretrained representations, but full fine-tuning can substantially change downstream EMG performance. Comparing each pretrained model with the same model trained from scratch shows that the benefit of pretraining varies substantially across models and is not universal. Performance decreases when models are evaluated on new users, while five-shot adaptation improves macro-F1 in 70.2% of evaluated model-dataset combinations but recovers only part of the lost performance. Model performance is highly consistent between upper- and lower-limb classification and remains strongly correlated with continuous EMG-to-text decoding. Together, these results provide a systematic view of when pretrained time-series models transfer effectively to EMG and how their performance depends on fine-tuning, user variation, and downstream task.

cs.LG↗
arXiv2026

LiteEMG-FM: An Efficient and Deployable Foundation Model for Robust EMG Sensing

Electromyography (EMG) signals vary substantially across individuals, body regions, recording sessions, and sensing hardware, limiting the generalization of models for assistive devices and human-computer interaction. Existing time-series foundation models are also computationally expensive for real-time wearable deployment and often fail to capture EMG-specific time-frequency characteristics. We present LiteEMG-FM, an efficient hybrid CNN-Transformer foundation model for practical EMG sensing. Pretrained on 16 diverse upper- and lower-limb EMG datasets, LiteEMG-FM learns representations that generalize across users and datasets. For resource-constrained deployment, we implement a hierarchical wake-up architecture in which a lightweight, always-on 1D-CNN filters rest and non-target activity and activates LiteEMG-FM only for valid gestures. We evaluate full inference offloading, split inference, and full on-device processing, characterizing their trade-offs in latency, power consumption, and memory footprint. Across diverse evaluation settings, LiteEMG-FM outperforms state-of-the-art time-series foundation models and supervised baselines, particularly under zero-calibration cross-participant and data-scarce conditions. These results demonstrate that LiteEMG-FM is an effective, efficient, and deployable foundation model for EMG applications.

cs.LG↗
arXiv2026

Scenario MPC with STL Specifications and Pareto-Based Feasibility Repair

Temporal logic is a formal language for reasoning about system behaviors over time. Signal temporal logic (STL), in particular, has been used to encode spatio-temporal requirements for control synthesis in multi-agent systems, often under the assumption that agents are cooperative and their dynamics are known. However, real-world multi-agent applications, such as autonomous driving, typically involve stochastic and uncontrollable agents. Recent work explored robust control with worst-case or probabilistic formulations, but remains limited in that it either (1) certifies strict satisfaction of STL constraints without addressing feasibility recovery, or (2) relaxes infeasible constraints with ego-centric objectives. In this paper, we propose a model predictive control (MPC) framework that treats feasibility repair as a Pareto optimization problem to explicitly characterize tradeoffs among agent objectives. We further provide a probabilistic certificate on STL violation rate to formally quantify uncertainty under stochastic and uncontrollable agents. The proposed framework is evaluated on two autonomous driving scenarios. Results show that the framework recovers feasible control with demonstrated safe behaviors.

cs.RO↗
arXiv2026

STAR: Sparse Tactile Representation Learning in Vision-Tactile-Language-Action Models for Dexterous Manipulation

Dexterous manipulation requires coordinated multi-finger control and effective tactile feedback, yet learning these capabilities remains challenging due to the lack of large-scale real-world data and the difficulty of extracting effective representations from sparse tactile signals. We build a robot platform and teleoperation system to collect a 200-hour bimanual dexterous manipulation dataset with synchronized visual, tactile, and language annotations, comprising 10,576 trajectories across 65 tasks, 69.5% of which involve dexterous multi-finger manipulation. We further propose STAR, an integrated training recipe for vision-tactile-language-action (VTLA) models that addresses the spatial, temporal, and informational sparsity of tactile signals through visual-tactile joint pre-training, sparse-global tactile token representation, and sparse future tactile prediction. Trained on this dataset, STAR achieves a 61% average success rate across four real-world tasks with 100 post-training trajectories per task, demonstrating dexterous performance under task-specific post-training.

cs.RO↗
arXiv2026

Exact Black Branes in General Dimensions in Higher-Curvature Scalar-Tensor Gravity

We obtain two families of neutral planar black branes in a potential-free dilaton-coupled Lovelock--Horndeski theory. Heterotic compactification relates higher-dimensional curvature to scalar dynamics and motivates the interaction structure. At fixed scalar zero mode, the horizon radius and temperature vary while the complete stationary Wald entropy remains constant and positive. The exponential scalar factor at the horizon exactly compensates the area growth, giving this behavior a direct geometric explanation. An exact coefficient analysis determines the two four-dimensional branches and their continuation to every higher dimension through closed coupling relations. The linear family is conformally flat and leaves the AdS scale free at fixed couplings; the fractional family is Weyl curved and fixes that scale. The common kinetic relation also excludes their realization through a real one-scalar flat-torus reduction of a metric--dilaton parent.

hep-th↗
arXiv2026

Exact Hairy Black Holes in Higher-Curvature Scalar--Tensor Gravity

We construct exact hairy black holes in Lovelock--Horndeski gravity and establish a relation between horizon topology, scalar hair and covariant charges. Rooted in the dimensional reduction of higher-curvature gravity, this framework connects scalar--tensor dynamics to the low-energy description of heterotic strings. In five dimensions, a single fixed theory admits a continuous planar family and two isolated hyperbolic black holes. The hyperbolic solutions share a locally AdS metric with distinct scalar dressings, while the planar family is Weyl curved and has a freely varying horizon scale. Both branches carry scalar hair regular on the future horizon. The hyperbolic construction extends above four dimensions; compatibility of the displayed planar profile selects five and seven dimensions. Time translations of the scalar act through a global shift, which enters the complete covariant charge balance. Along the planar family, a changing bulk shift-charge density supplies this balance despite zero radial flux and vanishing horizon charge variation. The hyperbolic roots admit no horizon-scale variation at fixed couplings.

gr-qc↗
arXiv2026

Holographic Renormalization of String-Derived Lovelock--Horndeski Theory

String-derived higher-curvature scalar--tensor gravities encode microscopic coupling data in boundary response, raising the question of whether holographic observables can reconstruct the underlying higher-dimensional parameters. We answer this question for the five-dimensional string-derived Lovelock--Horndeski (SDLH) theory on its exact linear-dilaton asymptotically locally AdS branch, constructing the renormalized generating functional for an arbitrary boundary metric and spacetime-dependent scalar source. A boundary-covariant radial hierarchy unifies the variational problem, local backreaction, logarithmic obstruction, finite one-point functions, and Ward identities. Two response determinants organize the recursion, resonant obstructions, and metric--scalar mixing. The Weyl anomaly condenses into an Euler density, a Weyl-squared density, and a single curvature--scalar square whose paired variations generate the metric and scalar obstructions. The resulting renormalized functional carries string-selected coupling data into boundary geometry, operator response, anomaly coefficients, and a calculable interface with gravitational observables. On the regular branch, four scalar-normalization-invariant holographic combinations admit a global rational inverse to the continuous reduced couplings. At fixed compactification dimension the map has maximal rank, while the curvature-anomaly sum reconstructs the higher-dimensional Gauss--Bonnet coefficient without sign ambiguity. Holographic response thus provides an explicit, overdetermined boundary fingerprint of the underlying string reduction.

hep-th↗
arXiv2026

Spectral Fractional Bosonic Strings: Exact Polyakov Measure and the Critical Dimension

Fractional Laplacians encode anomalous scaling and covariant nonlocal response in quantum matter and holographic boundary systems. We place the corresponding scalar spectral theory on a dynamical closed worldsheet. A metric-dependent intertwiner maps the fractional quadratic form to that of $D$ free bosons, while its zeta Jacobian changes the matter determinant weight from $D$ to $sD$. The prime-determinant transformation includes the zero-mode area factor and determines both the moduli density and the Weyl coefficient $c_{\rm det}=sD$. Weyl cancellation therefore selects $D_\ast=26/s$. On the integer critical locus $1\leq D\leq26$, $N=26-D$ spectator bosons turn the residual determinant into a local critical CFT. Its fixed-area genus-$g$ vacuum measure is that of 26 free bosons, its torus trace is modular invariant, its plumbing channels are state resolved, and its BRST charge is nilpotent. The $D$ distinguished coordinates remain spectral composites; their pullback yields the Koba--Nielsen amplitudes and massless vertex conditions. Target backgrounds coupled to this distinguished sector retain the ordinary one-loop tensor beta functions, with dilaton deficit $sD-26$ and the corresponding string-frame action. Worldsheet gravity thus turns the fractional exponent into a quantum-geometric consistency parameter.

hep-th↗
arXiv2026

Sharp Plucker Geometry for Three-Copy Werner Distillation

Whether negative-partial-transpose entanglement can remain undistillable is a longstanding problem in quantum information theory. We analyze the first unresolved three-copy Werner endpoint using a sharp dimension-free inequality for complementary partial traces and an optimal exterior-square inequality for orthonormal tripartite vectors. The latter identifies local SWAP- parity statistics with metric data of a decomposable Plucker bivector. Together these inequalities prove endpoint nonnegativity for every positive semidefinite rank-two coefficient operator and for the complete normal rank-two sector in arbitrary finite local dimensions. For genuinely nonnormal operators, an exact crossed-Gram criterion proves nonnegativity when one local outpu-input support overlap is at most two, including every system with a qubit-sized factor, and when either support plane contains a product ray. Explicit anti-state and rank-boundary families establish optimality of the constants and the rank restriction.

quant-ph↗
arXiv2026

Autodata: An agentic data scientist to create high quality synthetic data

We introduce Autodata, a general method that enables AI agents to act as data scientists who build high quality training and evaluation data. We show how to train (meta-optimize) such a data scientist agent, so that it learns to create even stronger data. We describe the overall formulation, and a specific practical implementation, Agentic Self-Instruct. We conduct experiments on computer science research tasks, legal reasoning tasks and reasoning with mathematical objects, where we obtain improved results compared to classical synthetic dataset creation methods. Further, meta-optimizing the data scientist agent itself delivers an even larger performance uplift. Agentic data creation provides a way to convert increased inference compute into higher quality model training. Overall, we believe this direction has the potential to change the way we build AI data.

cs.AI↗
arXiv2026

An Algebraic Framework for Quantitative Semantics of Spatio-Temporal Logic with Graph Operators

Spatio-Temporal Logic with Graph Operators (STL-GO) extends Signal Temporal Logic (STL) to multi-agent systems via graph operators that count neighboring agents satisfying a property, together with multi-agent quantifiers. While Boolean semantics for STL-GO are well-defined, quantitative semantics have not yet been developed and existing quantitative semantics for spatio-temporal logics such as STREL cannot capture the counting constraints in STL-GO's graph operators. We develop quantitative semantics for STL-GO as a layered algebraic construction that separates temporal aggregation from graph-operator aggregation (governed by an abstract accumulator with a monotone fold and readout). We prove that soundness and completeness reduce to monotonicity conditions on these components. We implement the framework and evaluate it on two multi-agent environments: a 2D bounded region with stochastic Dubins-car dynamics and a 3D Earth-satellite system, under four semantic instantiations (Boolean, min-max, signed-deficit, and a hybrid), demonstrating the tradeoffs between accumulator choices and reporting scalability in the number of agents and time horizon.

cs.LO↗
arXiv2026

Lyapunov-Based Sample Complexity Analysis for Weakly-Coupled MDPs

We study the sample complexity of learning in average-reward weakly-coupled Markov decision processes (WCMDPs) and Restless Bandits (RBs) under a generative model. Naive reduction to a tabular MDP leads to high complexity bounds as the state-action space is exponentially large in the number of arms $N$. By exploiting the weakly coupled structure, we show that near-optimal policies can be learned with sample and computational complexities that are polynomial in $N$. Specifically, we analyze the plug-in approach, which applies an efficient planning algorithm to an empirical model estimated from data. For fully heterogeneous WCMDPs, we establish the first finite-sample PAC guarantee with polynomial complexity and an $O(1/\sqrt{N})$ optimality gap. For homogeneous RBs, we further prove that a smaller optimality gap is achievable under mild structural assumptions. A primary technical contribution of our work is a novel Lyapunov-based analysis framework. Unlike classical approaches that rely on the difficult-to-control bias function, our framework uses an explicitly constructed Lyapunov function along with a drift transfer technique between the true and empirical models. A key step of independent interest in our framework is a fine-grained perturbation analysis for the underlying linear programming (LP) relaxation, which provides a general tool for analyzing LP-based policies and weakly-coupled systems.

cs.LG↗
arXiv2026

CityTrajBench: A Unified Benchmark for City-Scale Vehicle Trajectory Generation

Urban trajectory generation is a fundamental task for transportation simulation, urban planning, and mobility analytics. However, systematic comparison across trajectory generation methods remains difficult because existing studies often rely on different datasets, preprocessing pipelines, trajectory representations, and evaluation metrics. This fragmentation makes it unclear whether reported performance differences arise from the generation mechanism itself or from inconsistent experimental protocols. To address this issue, we present CityTrajBench, a unified benchmark framework and protocol for city-scale vehicle trajectory generation. CityTrajBench standardizes data ingestion, trajectory normalization, feature construction, model adaptation, map-aware post-processing, model selection, and multi-level evaluation under a common setting. It supports heterogeneous generators, including statistical baselines, VAE-based, GAN-based, diffusion-based, and flow-matching-based models, and evaluates them on three real-world urban trajectory datasets. The benchmark measures global spatial realism, trip-level distribution fidelity, trajectory-level geometric similarity, conditional mobility consistency, and efficiency. Experiments reveal clear trade-offs across model families: DiffTraj is strongest on trajectory-level geometric fidelity, DiffRNTraj is competitive on structure-sensitive global realism, and TrajFlow provides a strong balance across realism, quality, conditional consistency, and efficiency. Meanwhile, a simple Markov baseline remains competitive on coarse-grained trip and local-movement statistics. These findings show that urban trajectory generation quality is inherently multi-objective, that no single model dominates all criteria equally, and that CityTrajBench provides a reproducible benchmark protocol and testbed for future research on urban mobility generation.

cs.LG↗
arXiv2026

WOW-Seg: A Word-free Open World Segmentation Model

Open world image segmentation aims to achieve precise segmentation and semantic understanding of targets within images by addressing the infinitely open set of object categories encountered in the real world. However, traditional closed-set segmentation approaches struggle to adapt to complex open world scenarios, while foundation segmentation models such as SAM exhibit notable discrepancies between their strong segmentation capabilities and relatively weaker semantic understanding. To bridge these discrepancies, we propose WOW-Seg, a Word-free Open World Segmentation model for segmenting and recognizing objects from open-set categories. Specifically, WOW-Seg introduces a novel visual prompt module, Mask2Token, which transforms image masks into visual tokens and ensures their alignment with the VLLM feature space. Moreover, we introduce the Cascade Attention Mask to decouple information across different instances. This approach mitigates inter-instance interference, leading to a significant improvement in model performance. We further construct an open world region recognition test benchmark: the Region Recognition Dataset (RR-7K). With 7,662 classes, it represents the most extensive category-rich region recognition dataset to date. WOW-Seg attains strong results on the LVIS dataset, achieving a semantic similarity of 89.7 and a semantic IoU of 82.4. This performance surpasses the previous SOTA while using only one-eighth the parameter count. These results underscore the strong open world generalization capabilities of WOW-Seg. The code and related resources are available at https://github.com/AAwcAA/WOW-Seg-Meta.

cs.CV↗
arXiv2026

Feasibility Restoration under Conflicting STL Specifications with Pareto-Optimal Refinement

Signal Temporal Logic (STL) is expressive formal language that specifies spatio-temporal requirements in robotics. Its quantitative robustness semantics can be easily integrated with optimization-based control frameworks. However, STL specifications may become conflicting in real-world applications, where safety rules, traffic regulations, and task objectives can be cannot be satisfied together. In these situations, traditional STL-constrained Model Predictive Control (MPC) becomes infeasible and default to conservative behaviors such as freezing, which can largely increase risks in safety-critical scenarios. In this paper, we proposes a unified two-stage framework that first restores feasibility via minimal relaxation, then refine the feasible solution by formulating it as a value-aware multi-objective optimization problem. Using $\varepsilon$-constraint method, we approximate the Pareto front of the multi-objective optimization, which allows analysis of tradeoffs among competing objectives and counterfactual analysis of alternative actions. We demonstrate that the proposed approach avoids deadlock under conflicting STL specifications and enables interpretable decision-making in safety-critical applications by conducting a case study in autonomous driving.

cs.RO↗
arXiv2026

A unified foundational framework for knowledge injection and evaluation of Large Language Models in Combustion Science

To advance foundation Large Language Models (LLMs) for combustion science, this study presents the first end-to-end framework for developing domain-specialized models for the combustion community. The framework comprises an AI-ready multimodal knowledge base at the 3.5 billion-token scale, extracted from over 200,000 peer-reviewed articles, 8,000 theses and dissertations, and approximately 400,000 lines of combustion CFD code; a rigorous and largely automated evaluation benchmark (CombustionQA, 436 questions across eight subfields); and a three-stage knowledge-injection pathway that progresses from lightweight retrieval-augmented generation (RAG) to knowledge-graph-enhanced retrieval and continued pretraining. We first quantitatively validate Stage 1 (naive RAG) and find a hard ceiling: standard RAG accuracy peaks at 60%, far surpassing zero-shot performance (23%) yet well below the theoretical upper bound (87%). We further demonstrate that this stage's performance is severely constrained by context contamination. Consequently, building a domain foundation model requires structured knowledge graphs and continued pretraining (Stages 2 and 3).

cs.CL↗