arXiv subjects

Jiahao Jiang

Publications and source records attributed to Jiahao Jiang.

15 recordsLinked to original sources

arXiv2026

Towards a Mathematical Theory of Adaptive Memory: From Time-Varying to Responsive Fractional Brownian Motion

This work attempts to develop a systematic mathematical description for a class of stochastic processes whose local regularity adjusts dynamically in response to their own state. The investigation proceeds in three stages. First, we consider a Time-Varying Fractional Brownian Motion (TV-fBm) with a deterministic, Hölder-continuous Hurst exponent function, and conduct a further theoretical analysis of its properties and structural features. For this process, we establish a range of fundamental properties, including an exact variance scaling law, local increment asymptotic estimates, local non-determinism, large deviation asymptotics for local increments, and a covariance structure that admits a closed-form hypergeometric representation. These results provide useful insight into the local and global dependence structure of the process. Second, as a central component of this work, we define a new class of processes termed Responsive Fractional Brownian Motion (RfBm). In this construction, the Hurst exponent is governed by a Lipschitz-Hölder response function that depends on the process state itself, thereby introducing an intrinsic feedback mechanism between state and memory. We establish the well-posedness of this definition through local existence and uniqueness results, prove that the induced instantaneous scaling exponent possesses almost sure Hölder regularity, and analyze the associated cumulative memory processes along with their asymptotic convergence properties. The framework offers a reference mathematical foundation for exploring state-dependent memory mechanisms.

math.PR↗

arXiv2026

A Two-Parameter Memory-Weighted Velocity Operator for Time and State Variables: Foundations and Fundamental Properties

We introduce and analyze a memory-weighted velocity operator V_{α,\b{eta}} within the framework of operator theory, establishing rigorous theoretical results for systems with time-varying memory. The operator employs two independent continuous exponents α(t) and \b{eta}(t) that separately weight past state increments and elapsed time scaling. This decoupling mechanism is the main novelty of the work: it allows the two memory aspects to evolve independently, which is essential for systems where the influence of past states and the perception of time may change in qualitatively different ways. Such situations arise naturally in adaptive materials, non-stationary transport, and evolving memory processes. Motivated by systems across multiple physical contexts-such as viscoelastic materials with stress-dependent relaxation or anomalous transport with history-dependent characteristics-the framework addresses memory aspects that may evolve differently across disciplines. We establish the operator's foundational properties: an explicit integral representation, linearity, and continuous dependence on the memory exponents with respect to uniform convergence. Central to the analysis are weighted pointwise estimates revealing how the exponent difference \b{eta}(t)-α(t) modulates the operator, and we establish weighted boundedness estimates for this linear operator. These estimates exhibit a natural compensation mechanism between the two memory weightings.

math-ph↗

arXiv2026

InternVLA-A1.5: Unifying Understanding, Latent Foresight, and Action for Compositional Generalization

Unified models for robot manipulation aim to equip one policy with both the semantic priors of pretrained VLMs and the physical dynamics learned through future prediction. In practice, existing designs tend to erode the semantics of the pretrained backbone, suffer interference among heterogeneous objectives, and learn future prediction from scratch in pixel space, leaving the dynamics priors of pretrained video generators unexploited. We present InternVLA-A1.5, which builds the policy on a native VLM backbone that keeps training on VQA and subtask prediction, and attaches a lightweight unified expert for continuous action generation. Future prediction is recast as a latent-querying problem, where a small set of learnable foresight tokens condenses the task-relevant future into a compact latent code under the supervision of a frozen pretrained video generation model, so the policy inherits world-model dynamics priors without ever learning pixel-level generation. The video branch is discarded at inference, keeping real-time control. Pretrained on 1.2M robot episodes and 3M multimodal samples, InternVLA-A1.5 achieves the best overall results on all six simulation benchmarks. In the real world, the preserved semantics deliver the strongest compositional generalization on held-out instruction bindings, and the two designs together sustain long-horizon execution.

cs.RO↗

arXiv2026

WSA$_1$: a 3D-Centric World-Spatial-Action Model for Generalizable Robot Control

Recent advances in embodied AI have established robot foundation models (RFMs) as the dominant approach for generalist robotic systems to date. By leveraging imitation learning on extensive robot demonstrations, RFMs have achieved impressive capabilities in mapping visual observations and language instructions to continuous robotic actions. However, current RFMs lack an inherent ability to reason about physical dynamics and the causal effects of robot behaviors on the 3D physical world. This creates a fundamental mismatch between 2D-centric visual perception and 3D-centric embodied interaction, severely limiting the generalization ability of RFMs in real-world tasks.To address this gap, we present WSA$_1$, a novel RFM built upon proposed 3D-Centric World-Spatial-Action modeling paradigm. It not only learns 3D world-aware visual thought for future robot behaviors, but also models mutual constraints between 3D world state transitions and robotic actions to enhance behavior generalization. Notably, WSA$_1$ achieves highly data-efficient pre-training with 6k hours of expert demonstration data (only 1k hours from real robot), while delivering competitive manipulation performance (93% success rate) on RoboTwin2.0 simulation benchmark and achieving +20% average boosted performance over state-of-the-art RFMs on real-world robot control tasks. These results reveal that generalizable RFM can be attained without large-scale real robot data when paired with 3D-centric world-action joint modeling, which offers a practical and affordable pathway to generalist robotic systems.

cs.RO↗

arXiv2026

A Functional-Analytic Framework for Nonlinear Adaptive Memory: Hierarchical Kernels, State-Dependent Sensitivity, and Memory-Dependent Functionals

This work develops a systematic functional-analytic framework for nonlinear adaptive memory, where the influence of past events depends on both elapsed time and the state values along a trajectory. The framework comprises three hierarchical layers. First, memory kernels are classified into mathematically admissible, regular (uniformly bounded, normalized, Lipschitz), and generalized (bounded variation, possibly sign-changing) classes. Second, adaptive sensitivity functions Lambda(s, f(s)) are introduced, satisfying natural conditions; a concrete construction based on historical deviation accumulation interpolates continuously between instantaneous response and history-dependent sensitivity, with an explicit Lipschitz estimate ||Lambda_f - Lambda_g||_inf <= L_Lambda ||f - g||_inf. Third, an adaptive memory-dependent functional S_{kappa, Lambda}(f) = sup_{t in I} (|f(t)| + integral_0^t Lambda(s, f(s)) kappa(t-s) |f(s)| ds) and the associated set M_{kappa, Lambda}(I) = {f : S_{kappa, Lambda}(f) < infinity} are constructed. Fundamental properties of the framework are established, including absolute convergence, measurability, uniform boundedness, positive definiteness, and comparison with the classical supremum norm. It is shown that C(I) is strictly contained in M_{kappa, Lambda}(I), with discontinuous functions (e.g., indicator functions of subintervals) belonging to the set -- capturing abrupt signal changes such as on-off switching in nonlinear systems. When the maximum of |f| is attained in the interior of the interval, a strict inequality S_{kappa, Lambda}(f) > ||f||_inf is proved, demonstrating the nontrivial contribution of the memory component.

math.FA↗

arXiv2026

Scalable High-Precision Microfabrication on Various Lithography-Incompatible Substrates and Materials Enabled by Wafer-Scale Transfer Lithography of Commercial Photoresists

Photolithography conventionally requires flat, rigid and stable substrates, limiting its applications in flexible, curved, and transient electronics. In this study, a breakthrough approach is reported that employs a reversibly adhesion-switchable phase-changing polymer to universally transfer commercial photoresists onto previously inaccessible substrates, overcoming fundamental limitations of conventional photolithography.

physics.app-ph↗

arXiv2025

MiVLA: Towards Generalizable Vision-Language-Action Model with Human-Robot Mutual Imitation Pre-training

While leveraging abundant human videos and simulated robot data poses a scalable solution to the scarcity of real-world robot data, the generalization capability of existing vision-language-action models (VLAs) remains limited by mismatches in camera views, visual appearance, and embodiment morphologies. To overcome this limitation, we propose MiVLA, a generalizable VLA empowered by human-robot mutual imitation pre-training, which leverages inherent behavioral similarity between human hands and robotic arms to build a foundation of strong behavioral priors for both human actions and robotic control. Specifically, our method utilizes kinematic rules with left/right hand coordinate systems for bidirectional alignment between human and robot action spaces. Given human or simulated robot demonstrations, MiVLA is trained to forecast behavior trajectories for one embodiment, and imitate behaviors for another one unseen in the demonstration. Based on this mutual imitation, it integrates the behavioral fidelity of real-world human data with the manipulative diversity of simulated robot data into a unified model, thereby enhancing the generalization capability for downstream tasks. Extensive experiments conducted on both simulation and real-world platforms with three robots (ARX, PiPer and LocoMan), demonstrate that MiVLA achieves strong improved generalization capability, outperforming state-of-the-art VLAs (e.g., $\boldsymbolπ_{0}$, $\boldsymbolπ_{0.5}$ and H-RDT) by 25% in simulation, and 14% in real-world robot control tasks.

cs.RO↗

arXiv2025

Pseudo-Label Refinement for Robust Wheat Head Segmentation via Two-Stage Hybrid Training

This extended abstract details our solution for the Global Wheat Full Semantic Segmentation Competition. We developed a systematic self-training framework. This framework combines a two-stage hybrid training strategy with extensive data augmentation. Our core model is SegFormer with a Mix Transformer (MiT-B4) backbone. We employ an iterative teacher-student loop. This loop progressively refines model accuracy. It also maximizes data utilization. Our method achieved competitive performance. This was evident on both the Development and Testing Phase datasets.

cs.CV↗

arXiv2025

Towards Better IncomLDL: We Are Unaware of Hidden Labels in Advance

Label distribution learning (LDL) is a novel paradigm that describe the samples by label distribution of a sample. However, acquiring LDL dataset is costly and time-consuming, which leads to the birth of incomplete label distribution learning (IncomLDL). All the previous IncomLDL methods set the description degrees of "missing" labels in an instance to 0, but remains those of other labels unchanged. This setting is unrealistic because when certain labels are missing, the degrees of the remaining labels will increase accordingly. We fix this unrealistic setting in IncomLDL and raise a new problem: LDL with hidden labels (HidLDL), which aims to recover a complete label distribution from a real-world incomplete label distribution where certain labels in an instance are omitted during annotation. To solve this challenging problem, we discover the significance of proportional information of the observed labels and capture it by an innovative constraint to utilize it during the optimization process. We simultaneously use local feature similarity and the global low-rank structure to reveal the mysterious veil of hidden labels. Moreover, we theoretically give the recovery bound of our method, proving the feasibility of our method in learning from hidden labels. Extensive recovery and predictive experiments on various datasets prove the superiority of our method to state-of-the-art LDL and IncomLDL methods.

cs.LG↗

arXiv2025

Microassembly of Multi-Material and 3D Integration Enabled by Programmable and Universal High-Precision Micro-Transfer Printing

Micro-transfer printing is an assembly technology that enables large-scale integration of diverse materials and components from micro- to nano-scale. However, traditional micro-transfer printing technologies lack dynamic selectivity, limiting capabilities in sorting and repairing materials and components for effective yield management during large-scale manufacturing and integration processes. In this work, we introduce a dynamically programmable micro-transfer printing system utilizing a sharp phase-changing polymer and an independently addressable microheater array to modulate adhesion through localized heating. The system demonstrates dynamically programmable capabilities for selective transfer of various materials including semiconductors, polymers and metals, handling geometries from micro-scale chiplets to nanometer-thick films and micro-spheres. It also exhibits exceptional capabilities in 3D stacking and heterogeneous materials integration, significantly advancing the manufacturability of complex microsystems. As a demonstration, we successfully perform dynamically programmable transfer of microLED chips to create arbitrarily specified patterns, offering a promising solution to the challenges of mass transfer and pixel repair in microLED display manufacturing.

physics.app-ph↗

arXiv2025

SLIF-MR: Self-loop Iterative Fusion of Heterogeneous Auxiliary Information for Multimodal Recommendation

Knowledge graphs (KGs) and multimodal item information, which respectively capture relational and attribute features, play a crucial role in improving recommender system accuracy. Recent studies have attempted to integrate them via multimodal knowledge graphs (MKGs) to further enhance recommendation performance. However, existing methods typically freeze the MKG structure during training, which limits the full integration of structural information from heterogeneous graphs (e.g., KG and user-item interaction graph), and results in sub-optimal performance. To address this challenge, we propose a novel framework, termed Self-loop Iterative Fusion of Heterogeneous Auxiliary Information for Multimodal Recommendation (SLIF-MR), which leverages item representations from previous training epoch as feedback signals to dynamically optimize the heterogeneous graph structures composed of KG, multimodal item feature graph, and user-item interaction graph. Through this iterative fusion mechanism, both user and item representations are refined, thus improving the final recommendation performance. Specifically, based on the feedback item representations, SLIF-MR constructs an item-item correlation graph, then integrated into the establishment process of heterogeneous graphs as additional new structural information in a self-loop manner. Consequently, the internal structures of heterogeneous graphs are updated with the feedback item representations during training. Moreover, a semantic consistency learning strategy is proposed to align heterogeneous item representations across modalities. The experimental results show that SLIF-MR significantly outperforms existing methods, particularly in terms of accuracy and robustness.

cs.IR↗

arXiv2024

BEVNav: Robot Autonomous Navigation Via Spatial-Temporal Contrastive Learning in Bird's-Eye View

Goal-driven mobile robot navigation in map-less environments requires effective state representations for reliable decision-making. Inspired by the favorable properties of Bird's-Eye View (BEV) in point clouds for visual perception, this paper introduces a novel navigation approach named BEVNav. It employs deep reinforcement learning to learn BEV representations and enhance decision-making reliability. First, we propose a self-supervised spatial-temporal contrastive learning approach to learn BEV representations. Spatially, two randomly augmented views from a point cloud predict each other, enhancing spatial features. Temporally, we combine the current observation with consecutive frames' actions to predict future features, establishing the relationship between observation transitions and actions to capture temporal cues. Then, incorporating this spatial-temporal contrastive learning in the Soft Actor-Critic reinforcement learning framework, our BEVNav offers a superior navigation policy. Extensive experiments demonstrate BEVNav's robustness in environments with dense pedestrians, outperforming state-of-the-art methods across multiple benchmarks. \rev{The code will be made publicly available at https://github.com/LanrenzzzZ/BEVNav.

cs.RO↗

arXiv2024

Label Distribution Learning from Logical Label

Label distribution learning (LDL) is an effective method to predict the label description degree (a.k.a. label distribution) of a sample. However, annotating label distribution (LD) for training samples is extremely costly. So recent studies often first use label enhancement (LE) to generate the estimated label distribution from the logical label and then apply external LDL algorithms on the recovered label distribution to predict the label distribution for unseen samples. But this step-wise manner overlooks the possible connections between LE and LDL. Moreover, the existing LE approaches may assign some description degrees to invalid labels. To solve the above problems, we propose a novel method to learn an LDL model directly from the logical label, which unifies LE and LDL into a joint model, and avoids the drawbacks of the previous LE methods. Extensive experiments on various datasets prove that the proposed approach can construct a reliable LDL model directly from the logical label, and produce more accurate label distribution than the state-of-the-art LE methods.

cs.LG↗

arXiv2019

Photonic hook generated by twin-ellipse microcylinder

Recently, two special photonic jets, photonic hooks and twin photonic jets, have been proposed to deal with complex conditions in nanoscale manipulation. Photonic hooks are generated by a single light plane wave and an asymmetric microparticle, while the twin photonic jets are produced by two incident light beams. In this letter, we presented and demonstrate a method to combine photonic hooks and twin photonic jets. A single light plane wave and a symmetric microparticle, twin-ellipse microcylinder, are used in this research. The curvature degree, length and maximum E2 filed enhancement of twin photonic hooks are varied significantly, with the change of refractive indices and shape of twin-ellipse microcylinder. And a liquid-immersed core-shell is built to achieve a flexible tunability.

physics.optics↗

arXiv2018

One-Pass Trajectory Simplification Using the Synchronous Euclidean Distance

Various mobile devices have been used to collect, store and transmit tremendous trajectory data, and it is known that raw trajectory data seriously wastes the storage, network band and computing resource. To attack this issue, one-pass line simplification (LS) algorithms have are been developed, by compressing data points in a trajectory to a set of continuous line segments. However, these algorithms adopt the perpendicular Euclidean distance, and none of them uses the synchronous Euclidean distance (SED), and cannot support spatio-temporal queries. To do this, we develop two one-pass error bounded trajectory simplification algorithms (CISED-S and CISED-W) using SED, based on a novel spatio-temporal cone intersection technique. Using four real-life trajectory datasets, we experimentally show that our approaches are both efficient and effective. In terms of running time, algorithms CISED-S and CISED-W are on average 3 times faster than SQUISH-E (the most efficient existing LS algorithm using SED). In terms of compression ratios, algorithms CISED-S and CISED-W are comparable with and 19.6% better than DPSED (the most effective existing LS algorithm using SED) on average, respectively, and are 21.1% and 42.4% better than SQUISH-E on average, respectively.

cs.DB↗