arXiv subjects

Huan Yang

Publications and source records attributed to Huan Yang.

At least 19 recordsLinked to original sources

arXiv2026

How Far Are We From On-Device VLM Inference? A Measurement Study on Edge NPUs

Diverse optimizations spanning model architecture, inference runtime, and hardware acceleration have been proposed for on-device vision-language model (VLM) deployment on edge NPUs. Still, cross-layer coordination of VLM execution on edge NPUs remains largely underexplored. This paper presents an empirical measurement study across six representative VLMs, three real-world edge platforms, and ten multimodal benchmarks. Our evaluation drills down from end-to-end application-level latency and thermal traces down to low-level kernel-scale execution behaviors, challenging several widely held community assumptions. First, softmax dominates attention costs; even lookup-table-based softmax optimizations cannot eliminate pipeline bubbles that severely underutilize NPU matrix units. Second, theoretical speedups from linear attention vanish under real-NPU execution due to prohibitive tensor-rearrangement overhead. Third, no hardware backend universally dominates autoregressive decoding: attention-tensor-shaped parallelism dictates NPU efficiency and shifts the CPU-NPU performance balance. We further find task-specific Sweet Resolutions for VLMs, which deliver meaningful accuracy gains at only modest latency overhead. Complementing these findings, we provide actionable cross-layer recommendations for hardware architects, runtime developers, and model designers.

cs.PF↗

arXiv2026

Gravitational-Wave Detector for Postmerger Neutron Stars: Beyond the Quantum Loss Limit of the Fabry-Perot-Michelson Interferometer

Advanced gravitational-wave detectors that have made groundbreaking discoveries are Michelson interferometers with resonating optical cavities as their arms. As light travels at finite speed, these cavities are optimal for enhancing signals at frequencies below their bandwidth frequency. A small amount of optical loss will, however, significantly impact the high-frequency signals which are not optimally amplified. We find an elegant interferometer configuration with an "L-resonator" as the core, significantly surpassing the loss limited sensitivity of dual recycled Fabry Perot Michelson interferometers at high frequencies. Following this concept, we provide a broadband design of a 25 km detector with outstanding sensitivity between 2-4 kHz. We have performed Monte-Carlo population studies of binary neutron star mergers, given the most recent merger rate from the GWTC-3 catalog and several representative neutron star equations of state. We find that the new interferometer configuration significantly outperforms other third-generation detectors by a factor of 3 to 7 in the signal-to-noise ratio of the post-merger signal. Assuming a detection threshold with signal-to-noise ratio >5 and for the cases we have explored, the new design is the only detector that confidently achieves a detection rate larger than one per year, with the rate being 1 to 30 events per year.

gr-qc↗

arXiv2026

CausGround: Causal Interaction Grounding for Adaptive Embodied Planning

Multimodal large language models (MLLMs) show promising reasoning capabilities, but reliable physical interaction requires planning that accounts for when actions are executable and how they change the world. These preconditions and effects are environment-dependent and often not captured by MLLMs, making adaptation challenging. Yet existing approaches either require costly environment-specific training or rely on unstructured interaction experience that is difficult to reuse. Therefore, we introduce CausGround, a non-parametric causal interaction grounding framework that distills interaction experience into reusable action-centered causal knowledge. CausGround identifies action preconditions and effects and organizes them in a factorized world-state representation capturing object attributes and inter-object relations. At planning time, task-relevant causal dependencies are grounded in current observations to support backward reasoning over preconditions and forward reasoning over effects for action selection and multi-step planning. We evaluate CausGround on simulated high-level planning, interaction reasoning, and real-world robot manipulation. On adapted VirtualHome and RoboCasa benchmarks, CausGround improves vision-based high-level planning across MLLM backbones, increasing average task success from 24.9% to 53.3% on VirtualHome and closed-loop planning success from 43.0% to 57.0% on RoboCasa. On SWITCH, average interaction reasoning accuracy increases from 25.8% to 31.9%, indicating improved action selection, recovery, and verification. Moreover, real-world experiments show that CausGround reasons backward from MLLM-interpreted goals to derive executable high-level action sequences for sequential manipulation. These results demonstrate that reusable action preconditions and effects support long-horizon embodied planning across diverse environments without parameter updates.

cs.RO↗

arXiv2026

A Broadband Room-Temperature Design for the Einstein Telescope

The unique underground location of the Einstein Telescope (ET) is expected to substantially suppress seismic and gravity-gradient noise, providing a favorable environment for gravitational-wave detection at low frequencies compared with surface-based observatories. The baseline ET design aims to achieve broadband sensitivity through a xylophone configuration, in which separate cryogenic and room-temperature interferometers are optimised for low- and high-frequency gravitational-wave detection. However, achieving the target low-frequency sensitivity introduces substantial technical complexity caused by vertical suspension thermal noise, vibrations of the cryocooler, and detuned optical configuration. In this paper, we demonstrate an alternative, entirely room-temperature, fully resonant, single-interferometer architecture that addresses these challenges and achieves broadband sensitivity superior to that of the conventional ET xylophone configuration. Using mature fused-silica suspension technology, recent advances in optical coatings, and novel inertial isolation, we show that 4-m-long fused-silica fibres can meet the required suspension thermal-noise performance. We further demonstrate that vibrations, optical rigidity, and quantum back-action noise can be controlled through appropriate design of the interferometer, test mass suspensions, optical seismometers, and quantum states.

gr-qc↗

arXiv2026

CoEvolve: Self-Evolving Code Policies from Skills to Strategies for Humanoid Robots

Recent advances in Vision-Language-Action (VLA) models and World Action Models (WAMs) have expanded the manipulation and whole body capabilities of individual robots. Yet a single robot remains limited by its embodiment, and many real world tasks require capabilities that can only emerge through cooperation. Existing multi robot frameworks largely emphasize task decomposition, allocation, and replanning, while providing limited support for adapting executable coordination when physical execution deviates from the intended behavior. We present CoEvolve, a framework that represents team coordination as executable code over fixed robot motor skills. CoEvolve introduces Verifier Guided Coordination Evolution, where a multimodal foundation model instantiates coordination strategies through motion primitives, semantic objectives, temporary robot roles, and bounded execution parameters. Physical verification evaluates each executed transition, rejected behaviors trigger structured coordination revision from an accepted checkpoint, and verified experience is retained to guide subsequent decisions. We evaluate CoEvolve with two simulated Unitree G1 humanoids transporting a freely supported rigid payload through constrained corridors. Across 15 base task runs with three multimodal backbone families, CoEvolve completes 8 tasks, with every backbone achieving end to end success. Additional evaluations yield verified placement on two additional corridor layouts and after both same-backbone and cross-backbone experience initialization, while exposing persistent limitations in motion precision and contact stability. These results support executable coordination as a practical interface for adapting long horizon multi robot collaboration under shared physical constraints.

cs.RO↗

arXiv2026

Modified Teukolsky Formalism for Extreme Mass-Ratio Inspirals in Higher-Derivative Gravity

In this work, we study a model problem involving a point particle spiraling into a non-rotating black hole in higher-derivative theories of gravity. In such theories, both the background spacetime and the generation and propagation of gravitational waves differ from those in General Relativity. We develop a modified Teukolsky formalism to describe gravitational waves sourced by the point particle and, as an illustrative example, compute the resulting fluxes to the black hole horizon and null infinity for a cubic gravity theory. The formalism is constructed in a way that can be naturally extended to rotating black holes. These results represent essential steps to build extreme mass-ratio-inspiral waveforms in modified gravity theories, which may also be rescaled to approximate waveforms from comparable-mass binary black hole systems, analogous to existing approaches in General Relativity.

gr-qc↗

arXiv2026

From the Test-Mass Limit to Binary Black-Hole Waveforms in Higher-Derivative Gravity

Many higher-derivative theories predict stronger deviations from General Relativity for lower-mass black holes, while their nonlinear field equations often prevent reliable simulations of the full binary evolution. Here we develop a route from controlled black hole perturbation theory based on the modified Teukolsky formalism to comparable-mass waveforms, using parity-even cubic gravity as a representative example. We find that the tidal response of the secondary black hole enters at the same perturbative order as the direct higher-curvature correction and is therefore essential for a consistent leading-order waveform. The resulting strong-field fluxes and conservative dynamics produce an accumulated inspiral dephasing that grows toward merger. Embedding this test-mass information into an effective-one-body model, we construct inspiral-merger-ringdown waveforms for comparable-mass binaries and find coupling-dependent dephasing and waveform-peak shifts. Our results demonstrate how strong-field test-mass calculations can anchor waveform models for higher-derivative gravity when theory-specific numerical-relativity simulations are unavailable.

gr-qc↗

arXiv2026

Modeling extreme mass-ratio inspirals in astrophysical and dark-matter environments

Extreme mass-ratio inspirals offer a powerful probe of massive black holes and their environments, but interpreting their signals requires accurate models of orbital dynamics and radiation. Here we bring together advances in black-hole perturbation theory to establish a unified relativistic framework for adiabatic inspirals in perturbative astrophysical environments. Built on the modified Teukolsky formalism, the framework requires neither post-Newtonian nor slow-rotation approximations. As a first application, we calculate the leading cloud-induced corrections to gravitational-wave fluxes and conservative orbital dynamics for circular, equatorial inspirals around a rotating black hole surrounded by a synchronized ultralight scalar cloud, combining these corrections with the previously calculated scalar response. In our fiducial system, a cloud containing $10^{-4}$ of the black-hole mass produces about $-57.3$ radians of orbital dephasing over two years. The phase contributions from gravitational-wave flux corrections and background-induced orbital shifts each exceed the scalar contribution in magnitude and partially cancel, highlighting the need for coupled matter-gravity modeling. Extensible to other environments and modified gravity theories, our framework provides the relativistic inputs needed to assess detection losses and parameter biases, and to probe the nature of black-hole environments with future space-based gravitational-wave observations.

gr-qc↗

arXiv2026

Cubic Quasinormal Modes in the Black-Hole Ringdown

We formulate and numerically solve the nonlinear Teukolsky equations through third order, obtaining a first-principles calculation of cubic quasinormal-mode amplitudes for Kerr black holes. Our approach combines lower-order metric reconstruction in split light-cone gauges with a hyperboloidal, compactified formulation solved on complex radial contours. Spin-dependent complex coupling coefficients fix the cubic amplitudes from products of their linear parent amplitudes. We apply these predictions to a nonprecessing numerical-relativity waveform, including cubic products of the $(2,2)$ fundamental mode, its first overtone, and their conjugates. The predicted cubic terms improve agreement over selected early-ringdown fitting windows and shift inferred linear-overtone amplitudes without adding fitting parameters. $3M$ after the peak of $|h_{22}|$, the summed cubic contribution reaches approximately $38\%$ of the total $h_{22}$ signal; $M$ is the initial total mass. These results identify an appreciable cubic response in early ringdown.

gr-qc↗

arXiv2026

Tidal Excitation of Relativistic Stars

How a relativistic star responds to a time-dependent tide has lacked a first-principles counterpart of the Newtonian normal-mode expansion. In general relativity, the tidal field is gauge sensitive, radiative modes are non-Hermitian and incomplete, and the direct and late-time responses are intrinsically nonmodal. In this work we present, to our knowledge, the first causal relativistic theory that treats all of these components on equal footing. We first develop the spectral decomposition of the retarded Green function of the matched stellar interior--Schwarzschild exterior, which classifies four separate regimes in time domain: acausal region, prompt wave, surface-reflected wave, and mode+tail. The prompt signal is identical to that of a Schwarzschild black hole with the same mass, which is combined with the surface-reflected wave to be referred as the "direct part". With the Green function and a complex bilinear form on the full metric--fluid phase space, we decompose the linear response of a nonrotating perfect-fluid star into direct, quasinormal-mode, and late-time-tail sectors. Crucially, every sector admits an explicit first-order driven evolution equation: discrete bilinear projections determine the dynamically evolving quasinormal-mode amplitudes, while continuum projections of the same tidal source onto branch-cut states determine the direct and tail amplitudes. The result is a complete modal--nonmodal decomposition of the spacetime and fluid perturbations, which is physically described by the dynamics of a discrete and a continuous-frequency set of oscillators. We verify the decomposition numerically under generic time-dependent tidal driving and discuss applications to black hole perturbations.

gr-qc↗

arXiv2026

ChunkTrust: Adapting Execution Horizons for Robot Policies with Action-Expert Evidence

Robot foundation policies predict action chunks, but how many actions to execute before replanning depends on the current task phase. We introduce ChunkTrust, which treats the execution horizon as a latent variable inferred from action-expert evidence rather than a fixed hyperparameter. Its training-free Action-aware Horizon Selector (AHS) combines intra-chunk spectral stability of generation traces with inter-chunk continuity between executed history and predicted actions. An online Beta posterior with kernel forgetting tracks horizon preferences across replans. A lightweight Query-based Horizon Adapter (QHA) optionally learns a context-conditioned dense prior from complementary evidence, fused with current evidence and episode-local Beta memory while the base policy remains frozen. Across RoboTwin2.0 and RoboCasa GR1 Tabletop, AHS improves overall task-averaged success for each evaluated base-policy configuration, including gains of +6.80 percentage points on $π_{0.5}$ over all 50 RoboTwin2.0 tasks and +9.67 percentage points on Qwen3GR00T in RoboCasa. AHS+QHA raises the gain over Base to +9.44 percentage points on the eight-task $π_{0.5}$ evaluation. On four real-world household tasks, AHS improves the equal-task mean normalized process score from 50.4% to 57.5%. Ablations examine the contributions of both evidence terms, temporal memory, and the learned prior. Project page is https://hf618.github.io/ChunkTrust.github.io/

cs.RO↗

arXiv2026

ROMA: LLM System for Real-World Object-Centric Multi-Sensory Active Perception

Humans inherently understand the physical world through an active process. When sensory evidence is insufficient to infer physical properties, we naturally interact with the environment by deciding what information is missing, how to acquire it, and when sufficient evidence has been obtained. In stark contrast, existing multi-sensory robot systems mainly integrate sensory inputs rather than actively acquiring missing evidence through interactions. In this work, we introduce ROMA, an LLM-based system for Real-World Object-Centric Multi-Sensory Active Perception. ROMA integrates vision, audio, tactile, and force sensing into a reasoning-interaction-feedback loop. The model identifies missing evidence and determines the target objects, interactions, and modalities, while a physical interface executes the selected interactions and collects the multi-sensory feedback. To support this capability, we construct ROMI-2K, a large-scale real-world multi-sensory object interaction dataset covering nearly 2,000 objects and 6 atomic interactions with synchronized sensory feedback. Building on these data, we develop a two-stage training framework that aligns sensory modalities and equips the LLM to assess evidence sufficiency, select informative interactions, and reason over the multi-sensory feedback. We further characterize active perception as perception chains, where acquired evidence guides subsequent interactions and reasoning, and establish ROMA Bench to evaluate single-attribute, long-horizon multi-attribute, and intent-driven active perception. Experiments show that ROMA can actively acquire missing evidence and solve complex, long-chain multi-sensory perception tasks that existing methods struggle to handle, laying a strong perceptual foundation for active multi-sensory embodied agents.

cs.RO↗

arXiv2026

EMBER-Bench: Benchmarking Cross-Event Causal Memory in Long-Horizon Embodied Tasks

Lifelong physical agents must reason over extended interactions where past events continue to shape the world long after they disappear from view. Beyond recalling what happened, agents must infer how history changes the current state and constrains future actions. Yet existing embodied and video-memory benchmarks largely focus on historical retrieval and summary, leaving such history-dependent causal reasoning underexplored. We introduce EMBER-Bench, an egocentric benchmark for cross-event causal reasoning in long-horizon embodied tasks, for which we newly created the task design, video recording, and data annotation. It contains 189 household tasks and 699 QA pairs, spanning task progress, failure recovery, external interventions, and compound long-horizon tasks with distant dependencies and prerequisites, with fine-grained event and causal-chain annotations. EMBER-Bench evaluates reasoning in both directions: next-action prediction selects the next action from history, and causal traceback, given that action, identifies the historical event that makes it necessary. Input ablations that add action logs or privileged cause-and-consequence annotations to the video indicate which kind of historical information models fail to use. Among the 16 evaluated models, the highest overall accuracy is 61.2%, compared with a mean of 98.3% across two human evaluators. At paired decision points, correct traceback is not associated with correct next-action prediction. Adding action logs yields a gain of 1.6 points, whereas cause-and-consequence annotations yield an additional gain of 13.0 points on top of that. These results suggest that extracting causal information from past events and converting it into constraints on current actions remains a key difficulty for long-horizon embodied agents. Project Page: https://zhaoalexgoat.github.io/EMBER-Bench/

cs.CV↗

arXiv2026

From Stars to Waves: Stochastic Inference of Microlensed Gravitational Waves

When strongly lensed gravitational waves pass through the stellar field of a lensing galaxy, they acquire additional phase and amplitude modulations from gravitational microlensing by stars and stellar remnants along the line of sight. These microlensed waveforms depend on the masses and positions of thousands or more of the most relevant stars required for numerical convergence of the wave-optics diffraction integral, so that their deterministic reconstruction from the data is computationally prohibitive. We formulate the detection and parameter estimation of such events as a stochastic inference problem and propose a solution with the implementation of normalizing flows. As a proof of principle, we use the inferred surrogate microlensing parameters to identify microlensing signatures and show that 10.2% of microlensed events can be detected at $\ge 3σ$ significance with third-generation gravitational-wave detector networks for source redshifts $z_s \le 2$. The correlation between the surrogate parameters and the density of the underlying stellar field further connects microlensing detection to the properties of the lensing galaxy. This approach opens the door to probing microlensing effects and the properties of the underlying stellar fields. Beyond microlensed gravitational waves, a similar construction may also be applied to other intrinsically stochastic inference problems, such as detecting post-merger gravitational waves from binary neutron star coalescence and signals from core-collapse supernovae.

gr-qc↗

arXiv2026

How significant is the lensing interpretation of GW231123?

GW231123 is one of the most unusual gravitational-wave (GW) events, with exceptionally large inferred masses and near-extremal spins, offering an opportunity to test whether propagation effects contribute to these properties. We therefore examine whether the data support wave-optics microlensing embedded in a strong-lensing galaxy, whose detection becomes increasingly likely as observations accumulate, whether this interpretation can explain these properties, and how significant the preference remains under detector noise and waveform systematics. We compare six hypotheses: unlensed, isolated point mass, and embedded point-mass (EPM) and binary-lens (EB) effective models in Type-I (minimum) and Type-II (saddle) macro images. The EB Type-I model is most favored. For the most accurate waveform model NRSur7dq4, it gives $\log_{10}B^{\rm EB-I}_{\rm U}=2.60$, versus $0.89$ for Type II, indicating sensitivity to macro-image geometry. Within Type I, however, the binary improves over the point mass by only $\log_{10}B^{\rm EB-I}_{\rm EPM-I}=0.16$ and $Δ\ln\mathcal{L}_{\max}=0.56$, providing no clear evidence for structure beyond a single effective perturber. Moreover, under embedded lensing, waveform-template discrepancies and inferred masses and spins are reduced. However, real O4a backgrounds from numerical-relativity injections show that the apparent lensing evidence is sensitive to waveform systematics and realistic detector noise: although the commonly used waveform IMRPhenomXPHM gives the largest Bayes factor, $\log_{10}B^{\rm EB-I}_{\rm U}=4.52$, it is less exceptional relative to its own background, with a false-alarm probability of $6.5$--$8\%$, whereas NRSur7dq4 gives only $2$--$3\%$. Thus, waveform systematics can amplify apparent lensing evidence, but GW231123 remains an intriguing lensing candidate.

astro-ph.GA↗

arXiv2026

Identifying Kilonovae in the Presence of Optical Afterglow for the Wide Field Survey Telescope

Identifying kilonovae associated with binary neutron star mergers is often complicated by the presence of a dominant synchrotron afterglow. In this work, we evaluate the performance of the Wide Field Survey Telescope (WFST) in identifying kilonova signals in composite afterglow-kilonova transients. Using a numerical framework based on the Fisher information matrix, we simulate $10,000$ realizations for each of two scenarios: an AT2017gfo-based template model and a physically sampled population that accounts for kilonova diversity. Our results indicate that kilonova identification is primarily limited by source distance. In both scenarios, the identification efficiency is largely insensitive to variations in afterglow microphysical parameters and exceeds $80\%$ at distances within approximately $600~\rm Mpc$ for AT2017gfo-like events. Under our adopted assumptions and a short gamma-ray burst (sGRB)-triggered target-of-opportunity (ToO) observational strategy, we estimate that the WFST could identify $0.1-1.2$ kilonovae per year in the optimistic scenario. Furthermore, we find that the discriminating power of color-based filters rapidly saturates, reaching a stable plateau by the second night after the merger. We therefore propose a staged observing strategy that prioritizes high-cadence $g$ and $r$-band monitoring during the first night and incorporates the $z$ band from the second night onward. This strategy improves the identification precision by exploiting the increasingly prominent red excess produced by the kilonova. Our results provide a physical basis for optimizing WFST observing resources to efficiently detect and characterize kilonovae in the multimessenger era.

astro-ph.HE↗

arXiv2026

A nonlinear voice from GW250114 ringdown

Gravitational-wave astronomy, by detecting ripples in spacetime, has opened a new window to observe compact objects and probe theories of gravity in the nonlinear strong-field regime. The ringdown signal of a binary black hole merger contains a superposition of damped sinusoids known as quasi-normal modes (Ref.[1]), whose frequencies are completely determined by the mass and spin of the remnant black hole and form the basis of black hole spectroscopy (Refs.[2-4]). A crucial prediction yet to be observationally confirmed is the existence of quadratic quasi-normal modes, which represent fundamental properties associated with wave-wave coupling in general relativity, and the leading mode is predicted to be detectable with next-generation ground-based detectors (Refs.[5-7]) using traditional methods. Here we show the first observational evidence for a set of quadratic quasi-normal modes in the ringdown of the binary black hole merger GW250114, the loudest gravitational-wave event detected to date, enabled by a novel analysis. These nonlinear modes result from the quadratic coupling of the linear $(2,2,n)$ modes with $n\leq3$. Starting the analysis at a time corresponding to four times the remnant mass ($M_\mathrm{f}$) after the merger, the evidence for their presence reaches a Bayes factor of 62. A phenomenological test allowing these modes to deviate from the theoretical prediction rejects the zero-amplitude hypothesis at a significance of 3.4 $σ$, while the inferred amplitude and complex frequency are consistent with the prediction of general relativity. This finding provides the first observational evidence of gravitational wave-wave interaction and extends black hole spectroscopy from the linear to the nonlinear regime. It also establishes a new direction for testing the fundamental nonlinear structure of general relativity with the most extreme gravity.

gr-qc↗

arXiv2026

Gravitational lensing of gravitational waves by galaxy clusters

Strongly lensed gravitational waves (GWs) are commonly searched for using population priors, including time-delay and magnification-ratio distributions, to suppress false alarms. Current LIGO, Virgo, and KAGRA (LVK) searches mainly adopt priors calibrated with galaxy-scale lenses; however, these may penalize the longer time delays produced by galaxy clusters, which could contribute 10%--20% of lensed events. Cluster member galaxies may also perturb the macro-lens potential and alter image magnifications. We investigate the time-delay and magnification-ratio distributions of binary black hole GWs lensed by galaxy clusters, together with microlensing systematics. In the full lensed sample, the median time delay is about 76 days for cluster lenses, compared with 6.9 days for galaxy lenses. After imposing an SNR threshold of $ρ>8$ at LVK O4 sensitivity, the medians are 15.4 and 0.24 days. The detected cluster sample retains a long-delay tail, with 9.1% of image pairs delayed by more than 365 days and 3.3% by more than 1000 days, so galaxy-based time-delay priors can suppress cluster-lensed pairs. By contrast, although cluster member galaxies broaden the magnification-ratio distribution of the full cluster sample, the detected cluster- and galaxy-lensing samples are very similar at O4 sensitivity, suggesting little bias in candidate selection. Cluster lenses also generally produce weak microlensing distortions: only about 0.8% of single-image mismatches exceed ${\cal M}=0.03$, while 4.8% exceed the parameter-estimation threshold $1.35/ρ^2$. Thus, microlensing is subdominant for most cluster-lensed events, mainly because images often form far from the brightest cluster galaxy, where the stellar density is low and dominated by intracluster light, although nearby satellites can locally enhance it. These results motivate a dedicated cluster-lensing pipeline for GW lensing searches.

astro-ph.GA↗