arXiv subjects

Haoxuan Wu

Publications and source records attributed to Haoxuan Wu.

9 recordsLinked to original sources

arXiv2026

Through the PRISM: Preference Representation in Intermediate States of Video Diffusion Models

Evaluating video generation with clean, pixel-based reward models disconnects evaluation from the noisy diffusion process and incurs massive VAE decoding costs. In this paper, we challenge this paradigm by asking a fundamental question: Can a powerful video generator inherently discriminate preferences directly from noisy latents? To answer this, we introduce \textbf{PRISM} (\textbf{P}reference \textbf{R}epresentation in \textbf{I}ntermediate \textbf{S}tates of Diffusion \textbf{M}odels). PRISM employs a lightweight Query-based Aggregation head with a frozen video diffusion backbone to decode preference signals from noisy latents. Surprisingly, PRISM not only achieves SOTA preference accuracy but also unlocks strong noise-robustness, which enables early-stage Best-of-$N$ sampling. This allows for filtering suboptimal candidates at the very beginning of denoising, drastically reducing computation while boosting video quality. We also reveal a strong positive correlation between a backbone's generative performance and its inherent evaluative power, enabling self-improving video backbones.

cs.CV↗

arXiv2026

MagCeptor: Encoding Broadcast-Addressable Logic into Magnetic Receptors

Multicellular coordination relies on broadcast-addressable receptors, yet engineered magnetic systems face an addressability bottleneck because global fields intrinsically conflate power and control. Here, we introduce MagCeptors to resolve this by encoding selectivity directly into magnetic topology. Establishing an energetic isomorphism with biological receptors, these arrays utilize local couplings to shape potential landscapes where global field vectors act as spatial keys, triggering deterministic snap-through instabilities. This architecture decouples force from source distance, achieving a density of 385 mN/mm3 (>50-fold increase over prior art). We validate this primitive through signal demultiplexing, embodied sequential logic, and untethered distributed networking. This framework enables distributed systems to orchestrate complex tasks without tethers or electronics, relying solely on the intrinsic logic of matter.

eess.SY↗

arXiv2026

Anatomical Landmark-Guided Deep Reinforcement Learning for Autonomous Gastric Navigation

Wireless capsule endoscopy (WCE) enables painless visualization of the gastrointestinal tract, but its diagnostic potential is limited by incomplete mucosal coverage and poor transferability of existing navigation methods across patient anatomies. We propose a transferable, anatomical landmarkguided deep reinforcement learning (AL-DRL) framework for autonomous gastric navigation. Leveraging a lightweight edgecontour-depth fusion module, our policy operates on stable, lowdimensional landmark coordinates rather than high-dimensional video streams, effectively bridging the sim-to-real gap. In simulations across eight patient-derived models, the method achieves over 97% coverage within 50 seconds, significantly outperforming vanilla PPO, SAC, and DQN agents. A two-stage sim-to-real pipeline with an adaptive dynamic programming controller actively mitigates physical disturbances. Ex-vivo experiments demonstrate a mean coverage of 87% and a 53% reduction in procedure time compared with expert manual control.

cs.RO↗

arXiv2024

SBoRA: Low-Rank Adaptation with Regional Weight Updates

This paper introduces Standard Basis LoRA (SBoRA), a novel parameter-efficient fine-tuning approach for Large Language Models that builds upon the pioneering works of Low-Rank Adaptation (LoRA) and Orthogonal Adaptation. SBoRA reduces the number of trainable parameters by half or doubles the rank with the similar number of trainable parameters as LoRA, while improving learning performance. By utilizing orthogonal standard basis vectors to initialize one of the low-rank matrices (either $\mathbf{A}$ or $\mathbf{B}$), SBoRA facilitates regional weight updates and memory-efficient fine-tuning. This results in two variants, SBoRA-FA and SBoRA-FB, where only one of the matrices is updated, leading to a sparse update matrix $\mathrmΔ \mathbf{W}$ with predominantly zero rows or columns. Consequently, most of the fine-tuned model's weights $(\mathbf{W}_0+\mathrmΔ \mathbf{W})$ remain unchanged from the pre-trained weights, akin to the modular organization of the human brain, which efficiently adapts to new tasks. Our empirical results demonstrate the superiority of SBoRA-FA over LoRA in various fine-tuning tasks, including commonsense reasoning and arithmetic reasoning. Furthermore, we evaluate the effectiveness of QSBoRA on quantized LLaMA models of varying scales, highlighting its potential for efficient adaptation to new tasks. Code is available at https://github.com/cityuhkai/SBoRA

cs.AI↗

arXiv2024

Trend and Variance Adaptive Bayesian Changepoint Analysis & Local Outlier Scoring

We adaptively estimate both changepoints and local outlier processes in a Bayesian dynamic linear model with global-local shrinkage priors in a novel model we call Adaptive Bayesian Changepoints with Outliers (ABCO). We utilize a state-space approach to identify a dynamic signal in the presence of outliers and measurement error with stochastic volatility. We find that global state equation parameters are inadequate for most real applications and we include local parameters to track noise at each time-step. This setup provides a flexible framework to detect unspecified changepoints in complex series, such as those with large interruptions in local trends, with robustness to outliers and heteroskedastic noise. Finally, we compare our algorithm against several alternatives to demonstrate its efficacy in diverse simulation scenarios and two empirical examples on the U.S. economy.

stat.ME↗

arXiv2024

Drift vs Shift: Decoupling Trends and Changepoint Analysis

We introduce a new approach for decoupling trends (drift) and changepoints (shifts) in time series. Our locally adaptive model-based approach for robustly decoupling combines Bayesian trend filtering and machine learning based regularization. An over-parameterized Bayesian dynamic linear model (DLM) is first applied to characterize drift. Then a weighted penalized likelihood estimator is paired with the estimated DLM posterior distribution to identify shifts. We show how Bayesian DLMs specified with so-called shrinkage priors can provide smooth estimates of underlying trends in the presence of complex noise components. However, their inability to shrink exactly to zero inhibits direct changepoint detection. In contrast, penalized likelihood methods are highly effective in locating changepoints. However, they require data with simple patterns in both signal and noise. The proposed decoupling approach combines the strengths of both, i.e. the flexibility of Bayesian DLMs with the hard thresholding property of penalized likelihood estimators, to provide changepoint analysis in complex, modern settings. The proposed framework is outlier robust and can identify a variety of changes, including in mean and slope. It is also easily extended for analysis of parameter shifts in time-varying parameter models like dynamic regressions. We illustrate the flexibility and contrast the performance and robustness of our approach with several alternative methods across a wide range of simulations and application examples.

stat.ME↗

arXiv2022

Syndrome decoding meets multiple instances

The NP-hard problem of decoding random linear codes is crucial to both coding theory and cryptography. In particular, this problem underpins the security of many code based post-quantum cryptographic schemes. The state-of-art algorithms for solving this problem are the information syndrome decoding algorithm and its advanced variants. In this work, we consider syndrome decoding in the multiple instances setting. Two strategies are applied for different scenarios. The first strategy is to solve all instances with the aid of the precomputation technique. We adjust the current framework and distinguish the offline phase and online phase to reduce the amortized complexity. Further, we discuss the impact on the concrete security of some post-quantum schemes. The second strategy is to solve one out of many instances. Adapting the analysis for some earlier algorithm, we discuss the effectiveness of using advanced variants and confirm a related folklore conjecture.

cs.IT↗

arXiv2022

Interpretable Latent Variables in Deep State Space Models

We introduce a new version of deep state-space models (DSSMs) that combines a recurrent neural network with a state-space framework to forecast time series data. The model estimates the observed series as functions of latent variables that evolve non-linearly through time. Due to the complexity and non-linearity inherent in DSSMs, previous works on DSSMs typically produced latent variables that are very difficult to interpret. Our paper focus on producing interpretable latent parameters with two key modifications. First, we simplify the predictive decoder by restricting the response variables to be a linear transformation of the latent variables plus some noise. Second, we utilize shrinkage priors on the latent variables to reduce redundancy and improve robustness. These changes make the latent variables much easier to understand and allow us to interpret the resulting latent variables as random effects in a linear mixed model. We show through two public benchmark datasets the resulting model improves forecasting performances.

stat.ML↗

arXiv2021

A News-based Machine Learning Model for Adaptive Asset Pricing

The paper proposes a new asset pricing model -- the News Embedding UMAP Selection (NEUS) model, to explain and predict the stock returns based on the financial news. Using a combination of various machine learning algorithms, we first derive a company embedding vector for each basis asset from the financial news. Then we obtain a collection of the basis assets based on their company embedding. After that for each stock, we select the basis assets to explain and predict the stock return with high-dimensional statistical methods. The new model is shown to have a significantly better fitting and prediction power than the Fama-French 5-factor model.

q-fin.ST↗