arXiv subjects

Haiyan Liu

Publications and source records attributed to Haiyan Liu.

At least 19 recordsLinked to original sources

arXiv2026

AR Forcing: Towards Long-Horizon Robot Navigation World Model

The diffusion based robot navigation world models are typically trained using parallel supervision, while autoregressive inference is employed during path planning. This results in a distribution shift between training and inference, which destabilizes the performance over long-horizon prediction. We propose AR Forcing, an autoregressive training strategy, which integrates the standard diffusion loss into the autoregressive training loop. At each step, the model uses its own predictions to update the context and optimize the single step noise prediction objective, thereby explicitly exposing the model to the inference state distribution during training. Our method does not require additional discriminators or distribution-matching losses, retains the original diffusion framework and sampler, and is easy to integrate. Experiments on multi-domain navigation datasets (RECON, SCAND, HuRoN, TartanDrive) show that compared with strong baselines, AR Forcing improved the consistency of generated images during long-horizon navigation and the accuracy of predicted trajectories, enhancing robustness of the model in complex known and unknown environments. We will release the code soon.

cs.RO↗

arXiv2026

Devil is in Narrow Policy: Unleashing Exploration in Driving VLA Models

We identify a fundamental Narrow Policy limitation undermining the performance of autonomous VLA models, where driving Imitation Learning (IL) tends to collapse exploration and limit the potential of subsequent Reinforcement Learning (RL) stages, which often saturate prematurely due to insufficient feedback diversity. Thereby, we propose Curious-VLA, a framework that alleviates the exploit-explore dilemma through a two-stage design. During IL, we introduce a Feasible Trajectory Expansion (FTE) strategy to generate multiple physically valid trajectories and a step-wise normalized trajectory representation to adapt this diverse data. In the RL stage, we present Adaptive Diversity-Aware Sampling (ADAS) that prioritizes high-diversity samples and introduce Spanning Driving Reward (SDR) with a focal style weighting to amplify reward's value span for improving sensitivity to driving quality. On the Navsim benchmark, Curious-VLA achieves SoTA results (PDMS 90.3, EPDMS 85.4) and a Best-of-N PDMS of 94.8, demonstrating its effectiveness in unlocking the exploratory potential of VLA models. Code: https://github.com/Mashiroln/curious_vla.git.

cs.CV↗

arXiv2026

PROSPECT: Unified Streaming Vision-Language Navigation via Semantic--Spatial Fusion and Latent Predictive Representation

Multimodal large language models (MLLMs) have advanced zero-shot end-to-end Vision-Language Navigation (VLN), yet robust navigation requires not only semantic understanding but also predictive modeling of environment dynamics and spatial structure. We propose PROSPECT, a unified streaming navigation agent that couples a streaming Vision-Language-Action (VLA) policy with latent predictive representation learning. PROSPECT uses CUT3R as a streaming 3D foundation spatial encoder to produce long-context, absolute-scale spatial features, and fuses them with SigLIP semantic features via cross-attention. During training, we introduce learnable stream query tokens that query the streaming context and predict next-step 2D and 3D latent features (rather than pixels or explicit modalities), supervised in the latent spaces of frozen SigLIP and CUT3R teachers. The predictive branch shapes internal representations without inference overhead. Experiments on VLN-CE benchmarks and real-robot deployment demonstrate state-of-the-art performance and improved long-horizon robustness under diverse lighting. We will release code for the community soon.

cs.CV↗

arXiv2026

From Representational Complementarity to Dual Systems: Synergizing VLM and Vision-Only Backbones for End-to-End Driving

Vision-Language-Action (VLA) driving augments end-to-end (E2E) planning with language-enabled visual backbones, yet it remains unclear how vision-language models (VLMs) differ from standard vision-only encoders and whether these differences survive policy learning. We study this question under a unified VLM-hidden + diffusion-policy paradigm, comparing multiple VLM families/scales (e.g., InternVL3 and Qwen3VL) with vision-only encoders (e.g., ResNet, ViT, and EVA-CLIP). We ask three questions: how similar are their representations, do residual differences induce meaningful behavioral differences, and how can they improve accuracy-cost trade-offs? We find that VLM and vision-only policies share a substantial common subspace after policy learning, yet both retain non-transferable residual subspaces. Using a Shared-Unique SAE, we show that these residual factors are behaviorally relevant: vision-only encoders are stronger in simple, geometry-dominant scenarios, whereas VLMs are stronger in long-tail, semantically complex, and interaction-heavy cases. The two policy families also exhibit distinct driving styles, with vision-only models being more conservative on average and VLMs more assertive. Exploiting the complementarity between a VLM branch and a ViT branch yields an oracle upper bound of 93.58 PDMS on NAVSIM. We introduce HybridDriveVLA, which runs both branches and uses a learned trajectory scorer for selection, improving PDMS to 92.10 (+1.30 over the VLM baseline), and DualDriveVLA, a fast-slow variant that invokes the VLM in only 15% of scenarios, achieving 91.00 PDMS (+0.20) with about a 1.9x latency speedup over the VLM baseline. Code will be released.

cs.RO↗

arXiv2025

SparseWorld: A Flexible, Adaptive, and Efficient 4D Occupancy World Model Powered by Sparse and Dynamic Queries

Semantic occupancy has emerged as a powerful representation in world models for its ability to capture rich spatial semantics. However, most existing occupancy world models rely on static and fixed embeddings or grids, which inherently limit the flexibility of perception. Moreover, their ``in-place classification" over grids exhibits a potential misalignment with the dynamic and continuous nature of real scenarios. In this paper, we propose SparseWorld, a novel 4D occupancy world model that is flexible, adaptive, and efficient, powered by sparse and dynamic queries. We propose a Range-Adaptive Perception module, in which learnable queries are modulated by the ego vehicle states and enriched with temporal-spatial associations to enable extended-range perception. To effectively capture the dynamics of the scene, we design a State-Conditioned Forecasting module, which replaces classification-based forecasting with regression-guided formulation, precisely aligning the dynamic queries with the continuity of the 4D environment. In addition, We specifically devise a Temporal-Aware Self-Scheduling training strategy to enable smooth and efficient training. Extensive experiments demonstrate that SparseWorld achieves state-of-the-art performance across perception, forecasting, and planning tasks. Comprehensive visualizations and ablation studies further validate the advantages of SparseWorld in terms of flexibility, adaptability, and efficiency.

cs.CV↗

arXiv2025

Functional principal component analysis for functional data with detection limits

When measurements fall below or above a detection threshold, the resulting data are missing not at random (MNAR), posing challenges for statistical analysis. For example, in longitudinal biomarker studies, observations may be subject to detection limits. Functional principal component analysis (FPCA) is commonly used method for dimension reduction of dense and sparse data measured along a continuum, but standard approaches typically ignore MNAR mechanisms by imputing detection limit values, leading to biased estimates of principal components and scores. Building on recent work by Liu and Houwing-Duistermaat (2022, 2023), who proposed estimators for the mean and covariance functions under detection limits, we extend FPCA to accommodate functional data affected by such limits. We derive the asymptotic properties of the resulting estimators and assess their performance through simulations, comparing them to standard methods. Finally, we illustrate our approach using longitudinal biomarker data subject to detection limits. Our method yields more accurate estimates of functional principal components and scores, enhancing the reliability of functional data analysis in the presence of detection limits.

stat.ME↗

arXiv2025

Distorted optimal transport

Classic optimal transport theory is formulated through minimizing the expected transport cost between two given distributions. We propose the framework of distorted optimal transport by minimizing a distorted expected cost, which is the cost under a non-linear expectation. This new formulation is motivated by concrete problems in decision theory, robust optimization, and risk management, and it has many distinct features compared to the classic theory. We choose simple cost functions and study different distortion functions and their implications on the optimal transport plan. We show that on the real line, the comonotonic coupling is optimal for the distorted optimal transport problem when the distortion function is convex and the cost function is submodular and monotone. Some forms of duality and uniqueness results are provided. For inverse-S-shaped distortion functions and linear cost, we obtain the unique form of optimal coupling for all marginal distributions, which turns out to have an interesting ``first comonotonic, then counter-monotonic" dependence structure; for S-shaped distortion functions a similar structure is obtained. Our results highlight several challenges and features in distorted optimal transport, offering a new mathematical bridge between the fields of probability, decision theory, and risk management.

math.OC↗

arXiv2024

Efficient Long-distance Latent Relation-aware Graph Neural Network for Multi-modal Emotion Recognition in Conversations

The task of multi-modal emotion recognition in conversation (MERC) aims to analyze the genuine emotional state of each utterance based on the multi-modal information in the conversation, which is crucial for conversation understanding. Existing methods focus on using graph neural networks (GNN) to model conversational relationships and capture contextual latent semantic relationships. However, due to the complexity of GNN, existing methods cannot efficiently capture the potential dependencies between long-distance utterances, which limits the performance of MERC. In this paper, we propose an Efficient Long-distance Latent Relation-aware Graph Neural Network (ELR-GNN) for multi-modal emotion recognition in conversations. Specifically, we first use pre-extracted text, video and audio features as input to Bi-LSTM to capture contextual semantic information and obtain low-level utterance features. Then, we use low-level utterance features to construct a conversational emotion interaction graph. To efficiently capture the potential dependencies between long-distance utterances, we use the dilated generalized forward push algorithm to precompute the emotional propagation between global utterances and design an emotional relation-aware operator to capture the potential semantic associations between different utterances. Furthermore, we combine early fusion and adaptive late fusion mechanisms to fuse latent dependency information between speaker relationship information and context. Finally, we obtain high-level discourse features and feed them into MLP for emotion prediction. Extensive experimental results show that ELR-GNN achieves state-of-the-art performance on the benchmark datasets IEMOCAP and MELD, with running times reduced by 52\% and 35\%, respectively.

cs.LG↗

arXiv2021

Image-free single-pixel segmentation

The existing segmentation techniques require high-fidelity images as input to perform semantic segmentation. Since the segmentation results contain most of edge information that is much less than the acquired images, the throughput gap leads to both hardware and software waste. In this letter, we report an image-free single-pixel segmentation technique. The technique combines structured illumination and single-pixel detection together, to efficiently samples and multiplexes scene's segmentation information into compressed one-dimensional measurements. The illumination patterns are optimized together with the subsequent reconstruction neural network, which directly infers segmentation maps from the single-pixel measurements. The end-to-end encoding-and-decoding learning framework enables optimized illumination with corresponding network, which provides both high acquisition and segmentation efficiency. Both simulation and experimental results validate that accurate segmentation can be achieved using two-order-of-magnitude less input data. When the sampling ratio is 1%, the Dice coefficient reaches above 80% and the pixel accuracy reaches above 96%. We envision that this image-free segmentation technique can be widely applied in various resource-limited platforms such as UAV and unmanned vehicle that require real-time sensing.

eess.IV↗

arXiv2021

AMA-GCN: Adaptive Multi-layer Aggregation Graph Convolutional Network for Disease Prediction

Recently, Graph Convolutional Networks (GCNs) have proven to be a powerful mean for Computer Aided Diagnosis (CADx). This approach requires building a population graph to aggregate structural information, where the graph adjacency matrix represents the relationship between nodes. Until now, this adjacency matrix is usually defined manually based on phenotypic information. In this paper, we propose an encoder that automatically selects the appropriate phenotypic measures according to their spatial distribution, and uses the text similarity awareness mechanism to calculate the edge weights between nodes. The encoder can automatically construct the population graph using phenotypic measures which have a positive impact on the final results, and further realizes the fusion of multimodal information. In addition, a novel graph convolution network architecture using multi-layer aggregation mechanism is proposed. The structure can obtain deep structure information while suppressing over-smooth, and increase the similarity between the same type of nodes. Experimental results on two databases show that our method can significantly improve the diagnostic accuracy for Autism spectrum disorder and breast cancer, indicating its universality in leveraging multimodal data for disease prediction.

cs.AI↗

arXiv2020

Social Network Mediation Analysis: a Latent Space Approach

Social networks contain data on both actor attributes and social connections among them. Such connections reflect the dependence among social actors, which is important for individual's mental health and social development. To investigate the potential mediation role of a social network, we propose a mediation model with a social network as a mediator. In the model, dependence among actors is accounted by a few mutually orthogonal latent dimensions. The scores on these dimensions are directly involved in the intervention process between an independent variable and a dependent variable. Because all the latent dimensions are equivalent in terms of their relationship to social networks, it is hardly to name them. The intervening effect through an individual dimension is thus of little practical interest. Therefore, we would rather focus on the mediation effect of a network. Although the scores are not unique, we rigorously articulate that the proposed network mediation effect is still well-defined. To estimate the model, we adopt a Bayesian estimation method. This modeling framework and the Bayesian estimation method is evaluated through a simulation study under representative conditions. Its usefulness is demonstrated through an empirical application to a college friendship network.

stat.ME↗

arXiv2019

Dynamic Dependence Modeling in financial time series

This paper explores the dependence modeling of financial assets in a dynamic way and its critical role in measuring risk. Two new methods, called Accelerated Moving Window method and Bottom-up method are proposed to detect the change of copula. The performance of these two methods together with Binary Segmentation \cite{vostrikova1981detection} and Moving Window method \cite{guegan2009forecasting} is compared based on simulated data. The best-performing method is applied to Standard \& Poor 500 and Nasdaq indices. Value-at-Risk and Expected Shortfall are computed from the dynamic and the static model respectively to illustrate the effectiveness of the best method as well as the importance of dynamic dependence modeling through backtesting.

q-fin.RM↗

arXiv2019

On estimation of the effect lag of predictors and prediction in functional linear model

We propose a functional linear model to predict a response using multiple functional and longitudinal predictors and to estimate the effect lags of predictors. The coefficient functions are written as the expansion of a basis system (e.g. functional principal components, splines), and the coefficients of the fixed basis functions are estimated via optimizing a penalization criterion. Then time lags are determined by simultaneously searching on a prior grid mesh based on minimization of prediction error criterion. Moreover, mathematical properties of the estimated parameters and predicted responses are studied and performance of the method is evaluated by extensive simulations.

stat.ME↗

arXiv2018

Prediction and forecasting models based on patient's history and biomarkers with application to Scleroderma disease

This paper aims at predicting lung function values based on patients historical lung function values and serum biomarkers in Scleroderma patients. The progression of disease is measured by three lung function indexes (FVC, TLC, DLCO). Values of four biomarkers (TIMP1, P3NP, HA, NT-proBNP) are available. The data are sparse (6 months intervals) and irregular (many visits are missed). We consider two modeling approaches to achieve our goal, namely, the mixed effects model which is the standard approach in epidemiological studies and the functional principal component analysis model which is typically used for dense temporal datasets. We find that functional data methodology was able to recover the trajectories of three lung function indexes and to predict the future values very well.

stat.AP↗

arXiv2018

On trend and its derivatives estimation in repeated time series with subordinated long-range dependent errors

For temporal regularly spaced datasets, a lot of methods are available and the properties of these methods are extensively investigated. Less research has been performed on irregular temporal datasets subject to measurement error with complex dependence structures, while this type of datasets is widely available. In this paper, the performance of kernel smoother for trend and its derivatives is considered under dependent measurement errors and irregularly spaced sampling scheme. The error processes are assumed to be subordinated Gaussian long memory processes and have varying marginal distributions. The functional central limit theorem for the estimators of trend and its derivatives are derived and bandwidth selection problem is addressed.

stat.ME↗

arXiv2014

Bayesian meta-analysis of correlation coefficients through power prior

To answer the call of introducing more Bayesian techniques to organizational research (e.g., Kruschke, Aguinis, & Joo, 2012; Zyphur & Oswald, 2013), we propose a Bayesian approach for meta-analysis with power prior in this article. The primary purpose of this method is to allow meta-analytic researchers to control the contribution of each individual study to an estimated overall effect size though power prior. This is due to the consideration that not all studies included in a meta-analysis should be viewed as equally reliable, and that by assigning more weights to reliable studies with power prior, researchers may obtain an overall effect size that reflects the population effect size more accurately. We use the relationship between high-performance work systems and financial performance as an example to illustrate how to apply this method to organizational research. We also provide free online software that can be used to conduct Bayesian meta-analysis proposed in this study. Research implications and future directions are discussed.

stat.ME↗

arXiv2010

Limiting shapes for a non-abelian sandpile growth model and related cellular automata

We present limiting shape results for a non-abelian variant of the abelian sandpile growth model (ASGM), some of which have no parallel in the ASGM. One of our limiting shapes is an octagon. In our model, mass spreads from the origin by the toppling rule in Zhang's sandpile model. Previously, several limiting shape results have been obtained for the ASGM using abelianness and monotonicity as main tools. As both properties fail for our model, we use a new proof technique: in our main proof, we introduce several cellular automata to mimic our growth model.

math.CO↗

arXiv2009

Existence and uniqueness of the stationary measure in the continuous Abelian sandpile

Let Λbe a finite subset of Z^d. We study the following sandpile model on Λ. The height at any given vertex x of Λis a positive real number, and additions are uniformly distributed on some interval [a,b], which is a subset of [0,1]. The threshold value is 1; when the height at a given vertex exceeds 1, it topples, that is, its height is reduced by 1, and the heights of all its neighbours in Λincrease by 1/2d. We first establish that the uniform measure μon the so called "allowed configurations" is invariant under the dynamics. When a < b, we show with coupling ideas that starting from any initial configuration of heights, the process converges in distribution to μ, which therefore is the unique invariant measure for the process. When a = b, that is, when the addition amount is non-random, and a is rational, it is still the case that μis the unique invariant probability measure, but in this case we use random ergodic theory to prove this; this proof proceeds in a very different way. Indeed, the coupling approach cannot work in this case since we also show the somewhat surprising fact that when a = b is rational, the process does not converge in distribution at all starting from any initial configuration.

math.PR↗