arXiv subjects

Yan Huang

Publications and source records attributed to Yan Huang.

At least 19 recordsLinked to original sources

arXiv2026

AWARE: Action-Entropy-Weighted and Attention-Guided Retrieval of Observation Evidence for Vision-Language-Action Models

Vision-Language-Action (VLA) models combine multimodal perception with robotic control, demonstrating remarkable potential for generalizable manipulation. Yet, enhancing their execution typically relies on auxiliary sensory inputs (e.g., depth, 3D point clouds) or costly fine-tuning. Inspired by the insight that Feed-Forward Networks (FFNs) act as key-value memory, we propose Action-Entropy-Weighted and Attention-Guided Retrieval of Observation Evidence (AWARE), an effective, training-free, and plug-and-play module for VLAs. Specifically, AWARE tracks layer-wise uncertainty via an element-wise Action Entropy metric, and retrieves observation evidence into the element's FFN through associative memory lookup. Crucially, the retrieved features are weighted by instruction-guided cross-modal attention to isolate task-relevant observation cues, while injection strength is adaptively scaled by the element's entropy and relative network depth. This selectively channels observation evidence to high-entropy action dimensions and deep layers where grounding decays, while leaving converged representations intact. Across LIBERO, SIMPLER, CALVIN, and real-world robot manipulation, AWARE consistently improves diverse VLA architectures with minimal computational overhead, achieving the largest gains on long-horizon tasks without architectural changes or additional training. Project page: https://jiabingyang01.github.io/AWARE/.

cs.CV↗

arXiv2026

Detection and characterisation of Starlink VHF TT&C emissions at 137.055 MHz with the 21 CentiMeter Array

Satellite emissions at low frequencies can interfere with astronomical observations. Observations with the EDA2 have revealed recurrent Starlink emissions near 137~MHz, attributed to very high frequency (VHF) telemetry, tracking, and command (TT\&C) transmissions. We aim to characterise Starlink VHF emissions detected with the 21 CentiMeter Array (21CMA) to better assess their potential impact on future low-frequency radio astronomical observations and inform the development of methods to mitigate the resulting interference. We searched five coordinated VHF channels in observations obtained on 15 April 2025 (UTC) and demodulated and decoded candidate signals. Two valid packets are detected at 137.055~MHz, each lasting 1.027584~s and separated by 99.987~s. A blind comparison with 11\,305 active satellites identifies STARLINK-32838 (NORAD ID 62827) as the transmitter, with differential Doppler measurements linking both events to the same spacecraft. Orbital analysis shows that the satellite was undergoing orbit raising and that the limited number of detections was partly attributable to its being below the horizon for most of the observing time. Re-analysis of EDA2 detections indicates that such emissions are not confined to orbit raising. The stronger event's estimated flux density is 9.23~MJy at 1000~km, consistent with previous measurements at the order-of-magnitude level. Signal demodulation and decoding, combined with orbital analysis, can improve the identification and characterisation of satellite-related radio frequency interference. The detection of similar VHF emissions across different orbital phases indicates that their potential interference with low-frequency radio astronomy should not be considered solely in the context of orbit raising.

astro-ph.IM↗

arXiv2026

A Point-source Catalogue from Direction-dependent Calibrated 21CMA Images

We present a deep low-frequency radio point-source catalogue centred on the North Celestial Pole (NCP) and extending $5^\circ$ in radius, derived from a single 24-h observation with the 21 CentiMeter Array (21CMA). Covering 75-175 MHz, the catalogue characterises extragalactic foreground point sources for Epoch of Reionization (EoR) studies. Previous 21CMA imaging has been limited by reduced dynamic range caused by direction-dependent effects, particularly the severe grating lobes of ultra-bright sources such as 3C 61.1 and ionospheric phase distortions. Here we apply 21CMAcali, a direction-dependent calibration and imaging pipeline, to mitigate these effects. The angular resolution ranges from 2.3 arcmin at 100 MHz to 1.4 arcmin at 168.75 MHz. The resulting maps show reduced artefacts and background noise, enabling more reliable recovery of faint sources. We construct a catalogue of 1246 radio point sources. Cross-matching with LoTSS-DR3, WENSS, and NVSS enables broad-band spectral fitting, while an independent in-band analysis yields a median spectral index of $α=-0.825$ (where $S \propto ν^α$). The Euclidean-normalised differential source counts at 150 MHz, derived from a dedicated 50-MHz-wide multi-frequency synthesis (MFS) image, are broadly consistent with existing measurements and semi-empirical models at the bright end, whereas the deficit at lower flux densities ($S \lesssim 0.04$ Jy) mainly reflects limited sensitivity and incompleteness. This catalogue improves foreground point-source characterisation in the 21CMA NCP field, demonstrates the value of direction-dependent calibration, and provides a reference for deeper catalogues based on multi-day observations.

astro-ph.IM↗

arXiv2026

Coupling Execution Horizon and Sampling Depth in Flow-Matching VLA Policies

Flow-matching Vision-Language-Action (VLA) models generate expressive continuous trajectories for robotic manipulation, yet balancing real-time responsiveness with execution stability remains a persistent challenge. Standard deployment workflows typically treat numerical solver depth and execution horizon as decoupled choices. The horizon amortizes inference latency, while solver depth is kept high to preserve integration fidelity. In this paper, we reveal a fundamental compute-execution coupling in flow-matching VLAs governed by a systematic sign reversal. While coarse numerical integration degrades precision under rapid replanning, it markedly improves task performance under extended open-loop commitment. Geometrically, numerical truncation contracts lateral action dispersion around an invariant trajectory centroid, serving as an implicit regularizer that suppresses compounding open-loop drift. Across LIBERO, RoboTwin 2.0, and real-world Franka manipulation on representative VLA backbones including $π_0$, $π_{0.5}$, and GR00T, exploiting this joint coupling consistently achieves superior task success while drastically lowering inference latency, entirely training-free and without architectural modifications. These findings establish numerical sampling depth as an operational variance regulator, shifting its role from a static precision budget to an active deployment knob in robot learning. Project page: https://jiabingyang01.github.io/Horizon-Depth-VLA/

cs.RO↗

arXiv2026

Relational Geometry Forcing for World Action Models

World Action Models (WAMs) have shown strong potential for robot manipulation by jointly learning visual dynamics and executable actions. However, their video and action objectives do not explicitly constrain the 3D geometry that is crucial for precise manipulation. Recently, geometry-aware WAMs introduce depth or 4D prediction branches, whereas representation alignment methods supervise intermediate WAM features with pretrained 3D foundation models. Yet we find that such pointwise alignment benefits the video features but degrades action learning, since dense geometric feature targets overly restrict compact action features. To address this limitation, we propose Relational Geometry Forcing (RGF), which transfers geometric priors from a frozen 3D foundation model by aligning pairwise feature relations instead of pointwise features. Specifically, RGF applies three complementary relational objectives, where intra-frame and cross-view relations transfer scene geometry and multi-view correspondence to the video features, while action-temporal relations couple action dynamics to geometric scene evolution. Applied only during training, RGF requires no explicit geometry branch or extra inference module. Extensive experiments in simulation and the real world demonstrate the effectiveness of Relational Geometry Forcing, showing consistent gains across diverse manipulation scenarios and WAM paradigms.

cs.RO↗

arXiv2026

OPIS: An Input-Grounded Benchmark for Multi-Object Memory in Video World Models

Video world models must preserve the visual state of the world over time, but existing evaluation protocols often rely on generated histories, video reference, or selected revisit viewpoints that can confound the assessment of a model's true memory capability. To address this, we introduce OPIS, an input-grounded benchmark that strictly anchors the assessment to a fixed set of object instances from the initial observation for evaluating multi-object memory in video world models. The OPIS dataset comprises 500 cases across real-world, embodied-robotic, and game-world domains, providing dense object-level annotations for 12,672 rigid, articulated, and deformable instances. Our object-centric evaluator combines association and explicit visibility reasoning to hierarchically measure Object (O) Presence (P), Identity (I), and Structure (S), utilizing static or dynamic evaluation tracks based on object kinematics. Across eight image-to-video or camera-conditioned world models, our proposed OPIS scores range from 48.65 to 56.01. As the reference inventory grows from less than 20 to more than 40 objects, the Presence, Identity, and Structure scores show an overall decline, with the average Identity score falling from 40.22 to 23.11. The results demonstrate that preserving the particular object instances in the input is considerably harder than generating plausible visual elements.

cs.CV↗

arXiv2026

Register Anything Model for Generalizable and Robust Point Cloud Registration

We present Register Anything Model (RAM), a unified model for generalizable 3D point cloud registration that robustly estimates 6-DoF transformations across diverse domains spanning sensors and environments. Existing registration methods are typically tailored to a single domain, with carefully tuned architectures and hyperparameters, and thus degrade markedly when applied to novel domains. We bridge this gap by training a single model on diverse data from indoor, outdoor, and object domains. A central challenge is the severe geometric discrepancy across domains, i.e., differences in scene scale, sensor noise, point density, and structural patterns, which hinders processing with a unified point-based model. To address this, we introduce a geometry-aware normalization strategy that jointly voxelizes and normalizes point clouds into a standardized geometric space, adaptively based on their geometric properties to preserve salient local structures while standardizing resolution. Built on this normalization, RAM follows a hierarchical coarse-to-fine paradigm, and employs an efficient geometric transformer that models local geometric consistency instead of globally over all points. This design mitigates discretization noise introduced by voxelization, and improves computational efficiency. Extensive experiments on 12 benchmarks demonstrate state-of-the-art performance in terms of inlier ratio and registration recall. Notably, our method outperforms the counterparts trained on each separate domain, showing superior generalization capability.

cs.CV↗

arXiv2026

When Seeing Is Not Believing -- A Benchmark for Search-Grounded Video Misinformation Detection

Video misinformation increasingly operates at the semantic and evidential level: authentic footage may be selectively edited, temporally reordered, spliced across sources, or augmented with AI-generated content to construct false narratives. Such evidence-dependent manipulations cannot be reliably verified from the input video alone, because the missing, reordered, replaced, or recontextualized evidence lies outside the video itself. We introduce \textbf{EVID-Bench}, a benchmark for search-grounded video misinformation detection, where a system must search the open web for related videos and identify what information is false through cross-video comparison. EVID-Bench comprises 222 videos spanning 9 manipulation types across 3 categories: AI generation, single-source editing, and multi-source editing. All samples are verified to be undetectable by frontier models through visual inspection alone. We evaluate nine frontier multimodal models using a retrieval-augmented verification baseline. The best system achieves only 61.43\% point-level accuracy and 43.24\% video-level accuracy, while AI-generated manipulations remain especially challenging. Error analysis reveals recurring challenges: models fixate on irrelevant anchors, misattribute synthetic content to editorial splicing, and terminate search prematurely before fully explaining the manipulation.

cs.CV↗

arXiv2026

What Matters for World Modeling in Vision-Language-Action Models?

Integrating world models into Vision-Language-Action models (VLAs) has emerged as a promising direction, enabling policies to anticipate future dynamics beyond reactive observation-to-action mapping. However, existing approaches vary simultaneously in their world representations, integration paradigms, and policy backbones, making it difficult to determine what truly matters for effective world model integration. To disentangle these factors, we categorize existing efforts into four representative paradigms, namely Implicit Future Alignment, Explicit Future Query, MoT Future Expert, and External World Predictor. We instantiate each within a unified framework and systematically evaluate these paradigms with diverse world representations and policy backbones across in-domain, out-of-domain, long-horizon, dynamic, few-shot, and real-world settings. Our study yields four main findings: 1) world representation effectiveness is paradigm-dependent, whereas 2D semantic representations offer robust, cross-paradigm gains; 2) no single paradigm is universally optimal, yet the decoupled MoT Future Expert offers competitive overall performance, particularly in OOD robustness; 3) world modeling consistently benefits policies built on standard VLMs, while decoupled designs exhibit greater compatibility with pretrained VLAs; and 4) world modeling improves performance over action-only baselines across multiple settings, with notable gains on long-horizon and dynamic tasks. Project page and video demonstrations are available at https://dylanmqs.github.io/WM-in-VLAs/.

cs.RO↗

arXiv2026

When Non-Adversarial Noise Hurts Learning: Benchmarking Skill Self-Evolution in Agents

Large language model agents increasingly improve by distilling experience into reusable skills, but errors in that experience can persist in learned rules and affect later tasks. Existing evaluations largely examine clean learning or perturbations encountered after skills are frozen, leaving robustness during self-evolution insufficiently understood. We introduce RSE-Bench, to our knowledge the first benchmark for evaluating skill self-evolution under ordinary, non-adversarial noise. It spans four domains and six benchmarks, with controlled perturbations to instructions, environmental evidence, execution records, and outcome--record pairings that preserve the underlying tasks and native scoring rules. The protocol separates noise during learning from input-side noise at evaluation, testing frozen skill stores on held-out tasks. We also introduce \textsc{RSEClaw}, a method-agnostic framework supporting consistent noise injection and process diagnostics of actual exposure and skill updates. Across three methods and two backbones, noisy learning retains only $25.5\%$--$42.1\%$ of the aggregate gain achieved by clean self-evolution, depending on the noise type. Perturbations confined to execution records and outcome--record pairings cause average losses comparable to input-side noise, despite leaving the original executions and evaluations intact. Diagnostics further identify differences in candidate skill text that coexist with small changes in mean held-out performance.

cs.LG↗

arXiv2026

Pion emission source shape in UrQMD Au + Au collisions at STAR energies

Femtoscopic measurements of two-pion Bose--Einstein correlations have established that particle-emitting sources in heavy-ion collisions are well described by Lévy $α$-stable distributions, motivating systematic studies across a wide range of collision energies. In this work, we present a three-dimensional femtoscopic analysis of pion pairs in Au+Au collisions simulated with the UrQMD model for collision energies $\sqrt{s_{NN}}=3$--$27\,\mathrm{GeV}$, taking the RHIC BES-II range of collider and fixed target experiment energies for reference. Using Lévy-type source parameterisations, we extract the Lévy index $α$, the three-dimensional radii $R_\mathrm{out}$, $R_\mathrm{side}$, $R_\mathrm{long}$, and the pair multiplicity parameter $λ^{*}$ (related to the correlation strength $λ$). We investigate their dependence on the transverse mass ($m_T$) and collision energy, along with derived quantities such as the radius difference $R_{\mathrm{diff}}^2=R_{\mathrm{out}}^2-R_\mathrm{side}^2$ and the ratio $R_\mathrm{out}/R_\mathrm{side}$. We find that $R_\mathrm{out,side,long}$ all decrease with increasing $m_T$ and increase with collision energy, consistent with collective expansion, $R_\mathrm{long}$ showing the strongest and $R_\mathrm{side}$ the weakest energy dependence. The Lévy index $α$ decreases with collision energy, with a larger $m_T$-dependence towards higher energies. The $λ^{*}$ parameter is consistent with a constant close to unity in the absence of pions from long-lived resonances. These results provide a baseline for future comparisons with experimental measurements from the STAR Collaboration, contributing to constraints on the QCD phase diagram.

nucl-th↗

arXiv2026

Femtoscopic parameters of pion emission sources in EPOS4 Au+Au collisions from 7.7 to 200 GeV

Three-dimensional femtoscopy probes the space-time structure of particle-emitting sources in relativistic heavy-ion collisions. The collision-energy dependence of Lévy source parameters may reveal the evolution of strongly interacting matter and possible signatures of the QCD phase diagram. We previously studied the excitation function of three-dimensional Lévy source parameters for pion pairs in Au+Au collisions with EPOS4, analyzing the pair-distance distribution $D(ρ)$ through one-dimensional projections. Here, we analyze the full three-dimensional source distribution and correlation function directly. We construct $D(ρ_{\rm out},ρ_{\rm side},ρ_{\rm long})$ as a 3D histogram, transform it into $C(q_{\rm out},q_{\rm side},q_{\rm long})$ using a Fast Fourier Transform, and fit the correlation function with the analytic Lévy formula. This approach can reduce biases arising when one-dimensional projections are used to describe a non-elliptical source. We apply the method across the RHIC Beam Energy Scan range and compare the extracted energy dependence with results from the projection-based method and recent UrQMD calculations. We focus on the out-side observables $R_{\rm diff}^{2}=R_{\rm out}^{2}-R_{\rm side}^{2}$ and $R_{\rm out}/R_{\rm side}$, whose different energy dependences in EPOS4 and UrQMD may illuminate the roles of emission duration and source dynamics. This work establishes a direct three-dimensional framework for comparison with future femtoscopic measurements and for exploring possible non-monotonic structures associated with the QCD phase diagram.

nucl-th↗

arXiv2026

EgoPhys: Estimating Peak Contact Force and Mechanical Work from Egocentric Manipulation Video

Physically grounded manipulation of articulated objects requires understanding both the maximum forces encountered during contact and the work performed as their parts move. Peak contact force and mechanical work quantify these complementary aspects, but estimating them from egocentric video is challenging because physical interaction cues are local and indirect. Moreover, peak force is associated with brief contact events, whereas mechanical work depends on force-motion coupling throughout the contact duration. To address these challenges, we propose EgoPhys, an RGB-only framework comprising Contact-Aware Spatial Aggregation (CASA) and Target-Specific Multi-Expert Temporal Routing (TMTR). CASA integrates appearance and geometry features to emphasize interaction-relevant cues, while TMTR models semantic, event, and motion cues with specialized temporal experts and routes them separately for force and work prediction. On the test split from Hoi! dataset, EgoPhys substantially improves predictions of peak force and mechanical work, achieving MAEs of $5.205 \pm 0.584$ $N$ and $0.894 \pm 0.081$ $J$, respectively.

cs.CV↗

arXiv2026

SymNetPro: LOS-Aware Directional Multi-Transmitter Localization from Sparse Radio Observations

Directional multi-transmitter localization from sparse received-power observations is difficult because the receiver observes only the source-unresolved aggregate field: multiple directional sources superpose, building blockage fragments their visible regions, and stronger sources can mask weaker ones. We present SymNetPro, which retains the dual-task radio-map reconstruction and localization backbone of SymNet and adds two targeted components. First, a sparse line-of-sight (LOS)-aware attention bias injects obstruction-aware spatial relations into selected token interactions. Second, transmitter-drop augmentation recomposes training scenes after removing one sample-supported transmitter, exposing the model to controlled source-cardinality variation. Experiments on directional ray-traced urban environments show substantially lower OSPA than representative localization baselines under extreme sparse sampling, with consistent gains under measurement noise and increasing transmitter count. A transmitter-specific evidence analysis further shows that remaining misses concentrate in regimes where the target contributes little distinguishable power to the aggregate observation.

cs.IT↗

arXiv2026

ViDAL: A Visual Dynamics-Grounded Action Latent Space for Vision-Language-Action Models

Vision-Language-Action (VLA) models have become a central paradigm for robot policy learning, which predict actions in three forms: raw action chunks, discrete action tokens, or continuous action latents. However, existing action representations primarily model action trajectories, with limited consideration of the visual dynamics induced by these actions. We introduce ViDAL, a Visual Dynamics-grounded Action Latent Space that anchors continuous action latents in the future visual dynamics of the scene. Specifically, ViDAL learns action latent space by training an Action Variational Autoencoder (Action VAE) to reconstruct action chunks while aligning its latent with future scene dynamics. When integrated into downstream robot policies, the proposed Action VAE serves as a plug-in action interface compatible with multiple VLA architectures and enables optional future-video prediction as an additional capability. Empirically, ViDAL outperforms competitive baselines on LIBERO with 98.1% average success, improves a multi-task $π_{0.5}$ policy on RoboTwin 2.0 from 54.3% to 65.5% (Clean) and from 33.2% to 43.1% (Random) success rates over 50 dual-arm tasks, and yields 20.0% and 23.4% absolute success-rate gains on real-world single-arm Franka and dual-arm ARX robot platforms.

cs.RO↗

arXiv2026

AlignQuant: Tile-Aligned Mixed-Precision Quantization for Efficient LLM Generation

Fine-grained mixed-precision quantization promises efficient large language model inference, but local precision choices can conflict with regular GPU storage and computation units. This precision-boundary mismatch limits the translation of compression into practical acceleration. We introduce AlignQuant, a post-training quantization method that uses GPU-compatible two-dimensional weight tiles as the common unit of precision allocation, compact storage, and execution. This shared partition lets precision follow sensitivity within output channels. Joint prefill/decode calibration scores precision reductions using projection-output perturbations weighted by language-model loss gradients under quantized activations. Phase-normalized scores prioritize higher precision for tiles important to either phase under a model-wide weight-storage budget. Each tile stores one selected representation, while phase-specialized kernels reuse the packed model and expand lower-bit weights for INT8 computation with 8-bit activations. Across four LLMs spanning 3B to 14B parameters, AlignQuant achieves up to $2.50\times$ generation speedup over BF16 while preserving model quality. Evaluations further cover three GPUs and contexts up to 64K tokens. These results show that local precision flexibility and regular GPU execution can coexist through a shared tile unit. The implementation is available at https://github.com/HanzhiZhang-Ulrica/AlignQuant.

cs.LG↗

arXiv2026

From bits to atoms: A survey of cross-layer safety in Embodied AI

Embodied Artificial Intelligence (AI) introduces cyber-physical risks because digital failures and attacks can cause physical harm. This survey organizes embodied AI safety through a four-layer reference architecture -- comprising the physical interaction, system and middleware, algorithm and decision, and cloud-edge and data layers -- and uses it to connect vulnerabilities, propagation paths, and safeguards across the system stack. It traces the evolution of safety paradigms and examines how faults and attacks cross architectural boundaries to produce physical consequences. Furthermore, the survey systematizes composite attack chains, outlines multi-layered defense strategies, and reviews representative embodied safety benchmarks. Ultimately, we advocate for unifying software alignment with rigorous hardware constraints to build trustworthy systems capable of operating safely within human-centric environments.

cs.RO↗

arXiv2026

Comparison of chisquare variants for Poisson-distributed data and ratios

Fits to binned, Poisson-distributed data are common in high-energy and heavy-ion physics, and they are particularly delicate in correlation function measurements, where the fitted observable is the ratio of two histograms. We compare several goodness-of-fit estimators on large toy data sets: the Neyman and Pearson $χ^2$, their Yates continuity-corrected versions, a Neyman $χ^2$ with the variance shifted by $1/2$, the Poisson log-likelihood, and the correlation function likelihood in which both the signal and the reference histogram are treated as Poisson distributed. For a single histogram with mean occupancy $λ$, the Neyman $χ^2$ underestimates the bin content by approximately one count, while the Pearson $χ^2$ overestimates it by approximately half a count; shifting the variance by $1/2$ changes the Neyman result only at order $1/λ$, and only the log-likelihood recovers the mean without bias. For the ratio of two histograms, Neyman-type estimators are biased by approximately $-3/λ$ in relative terms (3\% at $λ=100$), whereas the Pearson and likelihood-based fits are unbiased in the symmetric configuration studied here. The Yates correction leaves the fitted values essentially unchanged but systematically deflates the $χ^2$, resulting in unrealistically high confidence levels. Simple analytic expressions are derived that reproduce all observed biases. Since these biases do not decrease with the number of bins, whereas the statistical uncertainties do, they can dominate the uncertainty of high-statistics, finely binned measurements. We therefore recommend likelihood-based fits, in particular the correlation function likelihood, for femtoscopic and similar ratio analyses.

nucl-th↗