Searcharxiv⌕ Search

arXiv subjects

Xiaoyu Chen

Publications and source records attributed to Xiaoyu Chen.

At least 19 recordsLinked to original sources

TyphoonWorld: A Residual Token World Model for Probabilistic Forecasting of Typhoon Imagery

A satellite forecast of a typhoon should give a low-error picture of the next two days, together with an ensemble whose members are coherent fields and whose spread responds to the storm at hand. A small direct regression network already gives the lowest squared error on held-out storms, so a generative model has to earn its cost through the part of the future that the regression smooths away. We present TyphoonWorld, a discrete-token world model that generates only that part. Conditioned on tokens of the regression forecast, the latest frames and the best-track state at forecast time, it samples residual fields lead by lead, and members add these residuals to the regression center. We compare it under one protocol with per-pixel Gaussian networks, a diffusion model of the same residuals, residuals from other storms and an unconditional token world model, on 276 forecast windows from 46 held-out storms of Digital Typhoon. TyphoonWorld matches the squared error of the regression, lowers the CRPS of the area colder than 220 K, a direct measure of deep convection, by 14 to 39% against every other ensemble, and ties its spread to forecast error as closely as the Gaussian network, with a correlation of 0.78 across windows. It uses 56% of the parameters of the unconditional world model.

cs.CV↗

Herding with Multiple States: When the Prior Matters

In sequential observational learning with more than two states, the same information structure and preferences can yield learning for some full-support priors and herding for others. This disproves a conjecture of Kartik, Lee, Liu, and Rappoport (2024, working-paper version). In the counterexample, the cascade region has nonempty interior, yet learning occurs for generic full-support priors outside it. Focusing on information and preferences, we introduce two necessary conditions for learning. Escapability requires that signals can provide arbitrarily strong evidence that any given action is ex post suboptimal; pairwise escapability is its pairwise variant. If escapability fails, herding occurs for some initial prior; if pairwise escapability fails, it occurs for every full-support prior. For a class of environments with interval optimality and monotone likelihood ratios, escapability completely characterizes learning and herding independently of the full-support initial prior. With location-shift signals, this characterization reduces to pairwise distinguishability, a generalization of unbounded beliefs (Smith and Sørensen, 2000).

econ.TH↗

LFM4Ads: Large Foundation Model for Ads Recommendation

Recent work leverages large-scale pretrained foundation models (FMs) to transfer generalizable representations to downstream advertising tasks. Most existing approaches transfer user representations (URs) and item representations (IRs), while underutilizing the user-item cross representations (CRs) generated by deeper interaction layers. Although a recent work transfers target-aware embeddings (TAEs), it requires real-time upstream FM inference alongside the downstream, introducing additional engineering complexity and overhead. To address these limitations, we propose Large Foundation Model for Ads Recommendation (LFM4Ads), an efficient pretrain-then-transfer paradigm for cross-domain ad recommendation. LFM4Ads jointly exploits URs, IRs, and CRs from the upstream FM. It extracts sample-level CRs from pretrained interaction layers and aggregates them into user-level CRs using a time-aware exponential moving average, thereby decoupling upstream inference from downstream online serving. To use CRs efficiently, we propose three downstream usages at different levels of granularity: feature-level gating to fuse CRs into downstream embeddings, module-level fine-tuning to yield downstream CRs, and model-level retrieval by matching CRs with IRs. LFM4Ads has been deployed on Tencent's advertising platform since Q4 2024. It has supported 10+ launches across major scenarios, yielding a platform-wide gross merchandise value (GMV) lift of 2.45% and annual revenue gains in the hundreds of millions of dollars. Offline experiments on KuaiRand-1K further demonstrate our effectiveness. Our code is available at https://github.com/ZhuangQu/LFM4Ads.

cs.LG↗

Riemannian Ordinary Least Squares

Many scientific responses take values in Riemannian manifolds, with predictors that are scalar- and/or manifold-valued. A regression analysis in such settings must provide accurate prediction with hypothesis testing for individual coefficients. Yet no current manifold-regression framework accommodates both predictor types and enables per-coefficient hypothesis tests. We propose Riemannian Ordinary Least Squares (ROLS) as a tangent space conditional-mean model, which is the ordinary least squares (OLS) analogue for Riemannian manifolds. ROLS is well-defined under an explicit injectivity condition that keeps the logarithm maps single-valued, with cut-locus conditions for branch-sensitive cases outside it. Within the local geometric setting it admits a closed-form estimator, an $O(n^{-1})$ excess prediction-error bound with explicit curvature bias, and asymptotic $t$- and Wald tests for individual coefficients. With synthetic simulations and three real-data case studies, ROLS is shown to be competitive with the state-of-the-art baselines. In addition, the per-coefficient tests agree with effects reported in literature. We envision ROLS to be a standard tool for manifold-valued regression with coefficient-level inference to support scientific discovery, which can be interpreted in the same way as Euclidean OLS. The source code is available at https://github.com/XiaoyuChenPub/ROLS.

stat.ME↗

A Spectral Local-to-Global Principle for Spin Systems on Graphs with Girth At Least Five

It is proved that, for every $δ\in (0,1)$, the Glauber dynamics for the uniform distribution on proper $q$-colorings is rapidly mixing when $q \geq (1+δ)Δ$ and the underlying graph has girth at least $5$ and maximum degree $Δ= Ω_δ(1)$. This result also extends to general multi-spin systems satisfying a $\textit{local spectral contraction}$ condition, including the anti-ferromagnetic Potts model with $q\geq (1+δ)(1-β)Δ$. These results are achieved by a new spectral local-to-global principle on graphs with girth at least five for general multi-spin systems, and a novel Fourier analysis for Glauber dynamics on a star. The main ideas behind all the proofs were developed through several rounds of interaction with GPT-5.6 Sol Ultra.

cs.DS↗

An FPRAS for Counting Common Bases of Two Matroids

We design the first fully polynomial-time randomized approximation scheme ($\mathsf{FPRAS}$) for approximately counting and uniformly sampling common bases of two matroids given by their independence oracles. Moreover, our algorithms generalize far beyond this to Hadamard products of two probability measures on the Boolean cube satisfying a simple nonnegative curvature condition. These algorithmic primitives have myriad applications in statistical physics, polyhedral combinatorics, the theory of quantum many-body systems, and beyond. The central difficulty is that the hard constraints imposed by the two input matroids (or the two input measures, more generally) obstruct traditional approaches based on local Markov chains. Relaxing these constraints can restore rapid mixing, but configurations that satisfy the original hard constraints may become exponentially rare if this is done naïvely. Our main contribution is a novel approach for approximate counting that marries Markov chain Monte Carlo (MCMC) with convex programming to overcome these obstacles: - We relax the intersection by imposing an overlap penalty on the product measure formed by the two input measures. We prove, via an integrated Bochner-type method, that this "soft intersection" satisfies a Poincaré Inequality uniformly over all external fields. - We use samples from the relaxed distribution to solve a dual maximum entropy convex program inspired by the Gurvits capacity. This program reweights the soft intersection measure using external fields to boost the aggregate probability of configurations satisfying the hard intersection constraint, without changing their relative weights. We bound the success probability directly using the uniform Poincaré Inequality and smallness of the gradient norm. AI Disclosure: GPT-5.6 Sol Ultra and GPT-6 Astra Ultra were heavily used to develop the ideas in this paper.

cs.DS↗

An $\mathsf{FPRAS}$ for Mixed Volumes

We design the first fully polynomial-time randomized approximation scheme ($\mathsf{FPRAS}$) to estimate the mixed volume of $n$ arbitrary compact convex sets in $\mathbb{R}^{n}$ specified by weak membership oracles. This disproves a longstanding conjecture of Gurvits (2009). $\textbf{AI Disclosure}$ The main ideas were generated entirely by GPT 6 Astra. The purpose of this paper is to provide a human-written exposition of the algorithm.

cs.DS↗

Video Prediction Policy 2: Predict Better, Act Better

World action models (WAMs) have emerged as an important class of generalist robot policies, aiming to transfer video prediction priors to action learning. However, we find that existing WAMs frequently produce incorrect motion predictions in open-ended environment, leading to erroneous actions. We attribute this limitation to two factors: (1) base video models are not optimized for manipulation, and (2) naively incorporating action components into video models can substantially degrade their generalization capabilities. We introduce Video Prediction Policy 2 (VPP2), a WAM that enables strong zero-shot generalization in both video prediction and action generation. First, we curate a large-scale, diverse dataset of manipulation videos to continue pretraining the base video foundation model. We annotate video clips with detailed captions and perform \textit{event-level} video pretraining to promote generalization across open-ended manipulation tasks. Second, we post-train and distill the video model into a single-step visual planner with fixed prediction horizon. Finally, we introduce action module via a mixture-of-transformers (MoT) architecture to learn implicit inverse dynamics model. Experiments demonstrate three key results: (1) VPP2-14B outperforms Cosmos3-64B by 11.0\% points in video prediction instruction-following success rate on open-ended tasks; (2) VPP2 surpasses the strongest baseline by 18.5\% points in success rate on real-world zero-shot ALOHA manipulation tasks; and (3) following benchmark-specific post-training, VPP2 achieves the highest success rates among evaluated methods on the challenging LIBERO-Pro, LIBERO-OOD, and RoboDojo benchmarks.

cs.CV↗

Vorch-Omni: Multi-Task Orchestration of Sight and Sound

Recent advances in generative video modeling have enabled diverse generation, reference-based synthesis, extension, and editing, but existing approaches often rely on fragmented task-specific models. A general model must distinguish heterogeneous target, source, and reference signals to determine what to generate, preserve, or use as guidance, while reducing interference among tasks. Joint audio-visual generation further increases this challenge by introducing diverse conditioning and output configurations across modalities. We present Vorch-Omni, a unified multi-task framework for audio-visual synthesis based on an arbitrary-condition-to-arbitrary-output formulation. It flexibly treats video and audio signals as either conditioning inputs or generation targets. Token-level conditioning masks and task identifiers distinguish targets, source content, and references, while position types separate temporal context from independent conditions. To capture semantic and structural information, Vorch-Omni employs complementary visual conditioning pathways: a vision-language model interprets sampled frames with text instructions, and a video VAE encodes conditions into latent tokens for direct guidance. We further build a distributed data pipeline to curate diverse temporally aligned audio-visual clips, generate structured captions and metadata, and balance heterogeneous task distributions. Built on a single flow-matching diffusion transformer without task-specific architectural changes, Vorch-Omni supports over 10 tasks, including text-to-video, text-to-audio-video, image- and reference-conditioned generation, temporal extension, audio-driven generation, video transformation, and audio-visual editing. This unified framework provides a scalable foundation for general-purpose audio-visual generation and manipulation.

cs.CV↗

Vorch-IR: Long-Form Unified Multimodal Identity Replacement Video Generation

Video identity replacement seeks to transfer the identities of one or more subjects while preserving the motion, expressions, and temporal structure of a driving video. Existing methods largely target single-person settings and often require task-specific structural controls, such as masks or pose representations, limiting their flexibility in general multimodal editing systems. Progress on multi-person replacement is further constrained by the scarcity of paired training data. We present Vorch-IR, a unified framework that supports single- and dual-person identity replacement, with optional background replacement, in a single model. Built on LTX2, Vorch-IR jointly conditions on a driving video, indexed reference images, and a textual editing instruction. The reference images need not match the pose, layout, or spatial configuration of the driving video: their roles as subject or background references are specified through the instruction. Dense visual conditions are fused through self-attention, while a vision-language context establishes semantic correspondence through cross-attention. We further develop an automatic data construction pipeline that synthesizes paired supervision for all four editing settings. Experiments using automatic metrics and pairwise human evaluation demonstrate strong identity preservation, motion fidelity, and temporal coherence across diverse scenarios. A temporal overlapping inference strategy additionally extends the short-clip model to minute-long generation without autoregressive continuation.

cs.CV↗

An $\widetilde{O}(n^3)$-time FPRAS for the Permanent

We give an FPRAS for the permanent of an $n\times n$ $0/1$ matrix with running time $\widetilde O(n^3)$. Our algorithm extends to a strongly polynomial FPRAS for arbitrary nonnegative matrices with the same running-time bound (but in the unit-cost arithmetic model). Jerrum, Sinclair, and Vigoda (2004) gave the first FPRAS for the permanent of a nonnegative matrix. The running time was subsequently improved to $\widetilde{O}(n^7)$ by Bezáková, Štefankovič, Vazirani, and Vigoda (2008), and recently to $\widetilde{O}(n^6)$ by Chen, Vigoda, and Yang (2026). We introduce a multicommodity-flow bound inspired by electrical flows, replacing the usual path-length factor by routing energy. For the classical JSV chain with perfect-matching weights multiplied by $n^2$, we prove a relaxation-time bound of $O(n^3\log n)$ and obtain comparable bounds for estimating the stationary probabilities of having a specified pair of unmatched vertices. We use a separate chain on assignments to generate a warm start at any given temperature, avoiding the long initialization of the JSV chain. Finally, we replace the JSV chain with a new chain that uses only $n$ hole weights and speeds up hole-weight learning by a factor of $n$.

cs.DS↗

InstAgent: A Recursively Self-Improving Agent Operating System for Heterogeneous Robot Collaboration

Heterogeneous robot teams enable complex, long-horizon tasks by combining complementary sensing, mobility, and manipulation capabilities across multiple embodiments. Yet existing systems struggle to retain execution-time corrections as reusable experience, maintain task-relevant context memory across space and time for reliable handoffs, and explicitly represent the inter-robot dependencies required for structured collaboration. In order to address these challenges, we present InstAgent, a recursively self-improving (RSI) Agent Operating System for heterogeneous robot collaboration. InstAgent organizes team behaviors as agentic behavior trees (BTs) that explicitly model capability assignment, inter-robot dependencies and handoff checks. The RSI mechanism updates these collaboration programs from execution feedback and preserves successful revisions for reuse in future tasks, while shared multimodal spatio-temporal memory provides context for planning and handoff validation. Experiments on 20 real-world multi-robot tasks with three high-level models show consistent gains in task success, first-attempt completion, and execution efficiency.

cs.RO↗

SWE-Game: Can Coding Agents Build the Games We Want?

We introduce SWE-Game, a benchmark of 247 tasks grounded in 41 executable reference Godot games spanning 13 gameplay categories in 2D and 3D. Five task types cover development from a brief, implementation from a game design document, skeleton completion, repair of 83 injected-fault cases, and Godot-to-Unity porting. Reference materials specify the intended gameplay, while a shared instrumentation interface lets evaluator-owned drivers and probes execute actions and observe independently implemented games. Evaluation combines engine-state checks, certified reference-input replay, and agent-authored feature demonstrations to assess mechanic correctness, demonstrated playability, and behavioral restoration and preservation after repairs. Game-specific vision-language rubrics separately assess presentation. Across six models, Opus5 achieves the highest overall score in all five task types. Best overall scores remain below 60 out of 100 across the three construction tasks, with Brief-to-Game reaching 50.38. Analysis of reviewed submissions identifies requirement omissions and gameplay logic errors as predominant implementation problems. On human-labeled behaviors from 100 agent-built games, executable checks achieve 92.59% balanced accuracy, compared with 78.41% for a video-based VLM judge. Rubric-based visual scores reach a Spearman correlation of 0.829 with human ratings of 200 gameplay clips. Together, these results characterize current agent capabilities across game-development activities and support combining runtime evidence with visual assessment.

cs.AI↗

RealtimeWAM: How Fast Can I Run My World Action Model?

World Action Models (WAMs) combine visual dynamics modeling with action generation, but their high inference latency limits responsive robot control. Recent efforts accelerate inference by removing explicit future-video generation at test time, as in FastWAM, an approach that requires a specially tailored architectural design. More general caching strategies exploit feature redundancy, but redundancy alone does not capture the changing computational demands of closed-loop control. To address these challenges, we present RealtimeWAM, a general, training-free framework that coordinates parallel execution with adaptive computation for low-latency inference across diverse WAM architectures. We exploit layerwise dependencies to overlap observation processing with prediction. However, concurrent branches still compete for GPU resources, limiting the benefit of parallel execution. We therefore adapt computation throughout the pipeline through selective reuse, caching observation features in visually stable regions and reusing Transformer residuals while reserving additional refinement for small predicted adjustments. We evaluate RealtimeWAM on FastWAM and OpenWAM across RoboTwin, LIBERO, and LIBERO-Plus. On an RTX 4090, measured mean inference latencies are 24.09 and 63.09 ms, corresponding to average speedups of 8.90$\times$ and 10.67$\times$. Average success rates are 82.75% and 87.41%, respectively, within 0.02 and 0.53 percentage points of native inference. Across five real-world tasks, RealtimeWAM improves average success rates over native inference by 17.2 and 37.2 percentage points on FastWAM and OpenWAM, respectively.

cs.RO↗

Disk-Resident Graph ANN Search: An Experimental Evaluation

As data volumes grow while memory capacity remains limited, disk-resident graph-based approximate nearest neighbor (ANN) methods have become a practical alternative to memory-resident designs, shifting the bottleneck from computation to disk I/O. However, since their technical designs diverge widely across storage, layout, and execution paradigms, a systematic understanding of their fundamental performance trade-offs remains elusive. This paper presents a comprehensive experimental study of disk-resident graph-based ANN methods. First, we decompose such systems into five key technical components, i.e., storage strategy, disk layout, cache management, query execution, and update mechanism, and build a unified taxonomy of existing designs across these components. Second, we conduct fine-grained evaluations of representative strategies for each technical component to analyze the trade-offs in throughput, recall, and resource utilization. Third, we perform comprehensive end-to-end experiments and parameter-sensitivity analyses to evaluate overall system performance under diverse configurations. Fourth, our study reveals several non-obvious findings: (1) vector dimensionality fundamentally reshapes component effectiveness, necessitating dimension-aware design; (2) existing layout strategies exhibit surprisingly low I/O utilization (less than or equal to 15%); (3) page size critically affects feasibility and efficiency, with smaller pages preferred when layouts are carefully optimized; and (4) update strategies present clear workload-dependent trade-offs between in-place and out-of-place designs. Based on these findings, we derive practical guidelines for system design and configuration, and outline promising directions for future research.

cs.DB↗

Transfer-Stratified On-Policy Distillation for RL-Improved Reasoning Teachers

Reinforcement learning can substantially improve a reasoning teacher, but it is unclear which of those improvements survive when the teacher supervises a smaller on-policy student. We study this question in mathematical reasoning by comparing teacher lineages before and after GRPO, multiple student scales, direct GRPO, and several on-policy distillation objectives. The central finding is that transfer is structured rather than scalar: teacher strength alone does not make dense distillation competitive, while an RL-improved teacher creates useful but metric-dependent student gains. This motivates Transfer-Stratified On-Policy Distillation (TS-OPD), which screens training problems by the joint sampled success of the student and teacher, routes acquisition problems to gated forward KL, routes consolidation problems to gated reverse KL, and adds an entropy brake to protect sampled coverage. Across the main comparison, TS-OPD is the strongest student objective for macro average correctness with the GRPO-improved teacher, while pass@K remains more mixed. Ablations show that the gains come from routing and token gating rather than skipping problems. These results support a transfer-aware view of OPD: stronger teachers help when the supervision direction and token budget match the student's observed ability, not merely because the teacher endpoint is stronger.

cs.LG↗

BORA: Bridging Offline Reinforcement Learning and Online Residual Adaptation for Real-World Dexterous VLA Models

Vision-Language-Action (VLA) policies provide strong behavioral priors for dexterous manipulation, yet adapting them on real robots remains challenging because high-DoF contact failures are difficult for humans to correct and online interaction is expensive. We present BORA, an offline-to-online reinforcement learning system that integrates an action-conditioned critic into a consistency-policy VLA and reuses the learned critic for frozen-base residual adaptation. To obtain executable corrective data, BORA combines wearable arm--hand teleoperation with a demonstration-guided local policy that translates coarse human intent into coordinated, embodiment-specific finger motions for contact-rich skills. Online robot rollouts and human corrections are mixed with offline data to update only a lightweight residual actor, avoiding full-model fine-tuning. We evaluate BORA on six real-world tasks using single-arm and bimanual platforms equipped with two dexterous-hand models. With only 20 online trajectories per task, BORA improves average success from 60.8% to 82.5% on standard objects and from 52% to 70% on held-out objects, while policy assistance substantially improves intervention reliability in bimanual twisting. These results demonstrate a practical route from executable human correction to efficient real-robot VLA adaptation.

cs.RO↗

Beyond a single latent space: a dual-latent world model for long-horizon planning

Latent world models often struggle with long-horizon planning despite accurate short-term predictions. Recursive rollouts accumulate errors, while distance concentration in high-dimensional latent spaces can weaken goal discrimination. We introduce the Dual-Latent World Model (Dual-WM), which separates local execution and long-range planning through distinct state representations and dynamics models. The low-level model predicts action-conditioned transitions, while the high-level model uses learned macro-actions to plan over longer temporal spans. We also propose Long-Horizon Representation Learning with Weighted Rollout (LoRe), which supervises self-generated predictions at both levels. An analysis of recursive error propagation motivates exponential horizon weights with separate decay rates for the two temporal scales. During planning, the high-level model generates latent subgoals that the low-level model refines into actions for precise execution. We evaluate from-scratch Dual-WM on five goal-conditioned visual control tasks against the task-wise strongest baselines without actor-guided proposals. At goal offsets of 50 and 100 environment steps, mean success increases from 75.9% to 84.4% and from 61.4% to 69.5%, respectively. At offset 100, Dual-WM outperforms these baselines on all five tasks and improves mean success over LeWM by 30.8 percentage points. Ablations and supporting analyses provide evidence of more informative representations for goal evaluation and greater consistency under recursive prediction. These results highlight the value of separating temporal roles and training across multiple horizons for reliable latent planning. Our core implementation is available at https://github.com/DeLin1001/Dual-WM-Official.

cs.AI↗