Search arXiv⌕ Search

arXiv subjects

Yan Wang

Publications and source records attributed to Yan Wang.

At least 37 records · Page 2Linked to original sources
arXiv2026

SkipWorld: Latent Skip Planning with Distilled World Models for Efficient Visual Navigation

World-model-based visual navigation enables action planning by predicting future states under candidate trajectories. However, long-horizon rollout remains challenging: models are often trained with ground-truth visual histories but deployed in closed-loop prediction, and evaluating many candidates with step-by-step rollout is computationally expensive. We propose SkipWorld, a latent skip planning framework for efficient visual navigation. SkipWorld introduces three components: (i) Latent AutoRoll, which trains a world model to recursively roll out future states in latent space, reducing the mismatch between training and closed-loop inference; (ii) SkipDistill, which distills the latent autoregressive model into a skip world model that directly predicts long-stride future latent states; and (iii) Skip-then-Refine Planning, a dual-model coarse-to-fine planner that first filters candidate trajectories with the skip model and then reranks selected candidates with the stronger autoregressive model. This design preserves accurate trajectory evaluation while reducing redundant rollout computation. Experiments on RECON, HuRoN, SCAND, and TartanDrive show that, compared with NWM, SkipWorld reduces the average ATE by 4.2% and RPE by 2.7%, while decreasing planning time by 62.4% with a 2.7-fold speedup.

cs.RO↗
arXiv2026

Judges Follow the Reference: The Derivation Gap in Verifying Tool-Call Arguments

Tool-call judges evaluate argument values proposed by LLM agents. We find that they accept a correct value less often when their input holds no reference for it, that is, no text that says what the value should be. This happens when the value is a default value from the function schema that the user did not mention, when it has to be inferred, or when the text it comes from was cut from the judge input. We call this drop in acceptance the derivation gap. It appears across all seven judges we evaluate, including System One models CLM and Jev. The gap is not only a truncation artifact: correct default values have the largest gap on a benchmark with no truncation. A controlled source-relocation test shows that judges follow the reference in their input: they favor a visible incorrect value over a correct value whose source is hidden. We introduce DeRef (Derived Reference): a separate model reads the full history without seeing the candidate and writes a reference for the judge. Across three benchmarks, DeRef improves overall argument AUROC for all judges tested. On the controlled test, it raises AUROC from chance or below to between 98.40 and 99.59.

cs.AI↗
arXiv2026

A Proof of the Erdős Matching Conjecture for Large Matchings via Sharp Tail Probability for Sums of i.i.d. Nonnegative Random Variables

We resolve a conjecture of Łuczak, Mieczkowska and Šileikis on the maximal tail probability of sums of independent and identically distributed nonnegative random variables. Specifically, for every integer $k\ge1$ and every $0\le x\le1/k$, we prove \[ \sup_{\mathbb E X_1\le x}\mathbb{P}(X_1+\cdots+X_k\ge1) =\max\{1-(1-x)^k,(kx)^k\}, \] where the supremum is over nonnegative i.i.d.\ random variables $X_1,\ldots,X_k$. We also characterize all maximizing distributions. Using this result, we prove the Erdős Matching Conjecture for every fixed $k\ge3$, all sufficiently large $s$, and every $n\ge k(s+1)$. More specifically, we show that in this range, the maximum number of edges in an $n$-vertex $k$-uniform hypergraph containing no $s+1$ pairwise disjoint edges is \[ \max\left\{ \binom{k(s+1)-1}{k}, \binom nk-\binom{n-s}{k} \right\}. \] Moreover, every extremal family consists either of all $k$-sets meeting a fixed $s$-set or of all $k$-sets contained in a fixed $(k(s+1)-1)$-set.

math.CO↗
arXiv2026

Language2Drive: Do Vision-Language-Action Models Really Follow the Instructions in Autonomous Driving?

Recent Vision-Language-Action (VLA) models increasingly use natural-language instructions to condition autonomous driving behavior. Yet existing closed-loop benchmarks primarily measure route completion, safety, and rule compliance, leaving a fundamental question unresolved: does language actually determine which feasible behavior the vehicle executes? A trajectory may be safe and successful while still violating or even ignoring the passenger's instruction. To address this, we introduce \ltd, a benchmark designed to directly test this capability. Its key evaluation unit, the paired counterfactual unit (PCU), holds the initial driving scene fixed while changing only the instruction. This design enables us to distinguish isolated instruction success from counterfactual paired success across four instruction-following scopes. To test whether this distinction persists across evaluation regimes, we instantiate Language2Drive in complementary closed-loop CARLA simulation, reconstructed real-world scenes, and open-loop driving benchmarks, including our released CARLA-Language2Drive simulator with purpose-built maps. Evaluating representative driving VLAs across these settings reveals a striking mismatch between strong driving competence and weak language control: 93.2% of successful rollouts do not have a successful counterpart under the alternative feasible instruction. This failure is dominated by one-sided completion, where a model repeatedly executes one preferred behavior but fails when the same scene requires the counterfactual alternative. These results expose two fundamental limitations of current evaluation: isolated success and driving scores, which are widely used by conventional driving benchmarks, can substantially overstate language-following capability, while current driving VLAs themselves lack reliable instruction following under matched counterfactual tests.

cs.RO↗
arXiv2026

ChunkTrust: Adapting Execution Horizons for Robot Policies with Action-Expert Evidence

Robot foundation policies predict action chunks, but how many actions to execute before replanning depends on the current task phase. We introduce ChunkTrust, which treats the execution horizon as a latent variable inferred from action-expert evidence rather than a fixed hyperparameter. Its training-free Action-aware Horizon Selector (AHS) combines intra-chunk spectral stability of generation traces with inter-chunk continuity between executed history and predicted actions. An online Beta posterior with kernel forgetting tracks horizon preferences across replans. A lightweight Query-based Horizon Adapter (QHA) optionally learns a context-conditioned dense prior from complementary evidence, fused with current evidence and episode-local Beta memory while the base policy remains frozen. Across RoboTwin2.0 and RoboCasa GR1 Tabletop, AHS improves overall task-averaged success for each evaluated base-policy configuration, including gains of +6.80 percentage points on $π_{0.5}$ over all 50 RoboTwin2.0 tasks and +9.67 percentage points on Qwen3GR00T in RoboCasa. AHS+QHA raises the gain over Base to +9.44 percentage points on the eight-task $π_{0.5}$ evaluation. On four real-world household tasks, AHS improves the equal-task mean normalized process score from 50.4% to 57.5%. Ablations examine the contributions of both evidence terms, temporal memory, and the learned prior. Project page is https://hf618.github.io/ChunkTrust.github.io/

cs.RO↗
arXiv2026

Thermally adaptive textile inspired by morpho butterfly for all-season comfort and visible aesthetics

The longstanding challenge of transitioning from static, appearance-limited passive daytime radiative cooling (PDRC) materials to systems that are both dynamically adaptive and aesthetically versatile in personal thermal management stems from the inherent compromise between color saturation and cooling power. Inspired by the Morpho butterfly, which decouples structural color from thermal function, we introduce a modular adaptive textile that dissolves this conflict. A dynamic thermochromic membrane autonomously switches solar reflectance from 0.6 (heating state) to 0.9 (cooling state), while an APC layer supplies angle-independent, high-saturation structural color independent of the thermal state and fully replaceable, enabling on-demand color rewriting. Consequently, outdoor tests show 5-7 °C surface-temperature reduction versus commercial colored fabrics under hot conditions, with no compromise in solar-heating performance under cold conditions. Complemented by hydrophobicity and breathability, this color-reconfigurable, energy-free approach offers a general pathway toward aesthetic and adaptive thermal comfort.

physics.optics↗
arXiv2026

TouchScale: 500 Hours of Human Vision and Touch for Visual-Tactile Learning

Large-scale egocentric human interaction data is becoming an important source of physical supervision for embodied learning, yet video alone leaves the contact and pressure that characterize physical interaction unrecorded. Recent visual-tactile datasets provide this missing supervision, but their synchronized tactile data remain far smaller in volume than human video. Moreover, the largest resources often merge recordings from different sensors or annotation procedures, which makes the effect of data scale difficult to isolate. We therefore introduce TouchScale, a 500-hour dataset of contact-rich human interaction recorded with a single unified wearable setup. Its approximately 2K predefined task descriptions span everyday activities and structured manipulation, and each recording temporally aligns egocentric RGB-D video with wrist RGB video and dense full-hand bimanual tactile measurements. Compared with prior tactile data, training on the full TouchScale raises zero-shot contact IoU on data from an unseen tactile sensor from 0.134 to 0.383. Pretraining a visual encoder on TouchScale also yields the highest action recognition accuracy on three benchmarks among the compared visual-tactile datasets. Used for visual-tactile mid-training of a robot policy, TouchScale improves the average real-world success rate across four contact-rich manipulation tasks from 22.5% to 57.5%. With the sensor and collection protocol held fixed, both zero-shot tactile prediction and robot success show an overall upward trend as more TouchScale data is used. These results suggest that human visual-tactile data collected at scale with consistent sensing benefits both perception and robot manipulation. We will publicly release TouchScale, including all synchronized visual-tactile recordings and reconstructed object models, to support future research on scalable visual-tactile learning.

cs.CV↗
arXiv2026

VeriFine: Scaling Verification for Self-Improvement in Embodied Reasoning

Self-improving policies continually expose new failure patterns, changing what their judges must be able to verify. However, current fixed judges constrain both optimization feedback and the discovery of useful training examples, limiting further self-improvement. This challenge is even more acute in embodied reasoning, where reliable evaluation must account for spatial grounding, causal reasoning, and safety-aware decision-making. We introduce VeriFine, an agent harness framework that scales verification through the co-evolution of the policy, training curriculum, and judge. The Policy Improvement Loop uses a rubric judge to diagnose recurring failures, construct an adaptive curriculum, and optimize the policy. When progress plateaus and verification becomes a bottleneck, the Judge Improvement Loop selectively queries human guidance on informative failure cases and refines the judge through coactive calibration, in which humans and agents resolve disagreements and converge toward the objective rubric of physical reasoning. The revised judge then guides the next stage of data selection and policy optimization. Experiments on driving and robot navigation tasks demonstrate continuous self-improvement in both policy and judge capability across reinforcement and supervised fine-tuning. These results show how scaling verification supports continuous self-improvement as policy failure patterns evolve.

cs.AI↗
arXiv2026

Grounding What Shapes the Plan: Rethinking Groundedness for Physical Intelligence in Autonomous Driving

Driving models increasingly ground reasoning in causal relations, spatial structure, perceptual evidence, and predicted futures. These advances make reasoning more faithful to the driving scene, but leave a fundamental question unresolved: what should groundedness mean when the model ultimately outputs an action? Correctly grounded reasoning does not, by itself, ensure desirable driving outcomes. We introduce GroundAct, which starts from a simple premise: driving unfolds through physical entities and their interactions. Entities therefore become the unit of grounding; a lightweight reference token keeps each selected entity's continuous state addressable through symbolic reasoning; and only the referenced entities' interactions with the evolving proposal correct the plan. The result is an explicit path from what reasoning grounds to what the plan does, which we call grounded planning. To assess its practical value, we evaluate GroundAct in both open- and closed-loop settings. GroundAct shows strong open-loop planning across normal, out-of-distribution, and safety-critical scenarios, with closed-loop results extending this evidence to driving in simulation.

cs.AI↗
arXiv2026

EvoCast: Reliable Autonomous Research Agents for Iterative Forecasting Architecture Evolution

Deep time-series forecasting models have rapidly diversified, yet adapting them to a specific task still requires extensive expert effort in model selection, mechanism diagnosis, architecture design, implementation, and evaluation. Existing AutoML methods are constrained by predefined search spaces, while general-purpose LLM research agents lack reliable control over experimental protocols and model promotion. We introduce EvoCast, a fully autonomous research-agent system for iterative forecasting architecture evolution. EvoCast first establishes and diagnoses a task-specific baseline through executed mechanism ablations, then generates evidence-grounded research directions from dataset characteristics, diagnostic results, prior rounds, and failure records. Its central design, cognition-authority separation, assigns open-ended hypothesis generation and code implementation to LLM agents, while deterministic program authorities control source-edit boundaries, canonical evaluation, and promotion decisions. Experimental outcomes are accumulated as evidence to guide subsequent rounds. Results show that EvoCast completes complex architecture modifications with higher implementation success and lower agent-side token/time cost, and develops task-specific architectures that outperform selected baselines, strong forecasting models, and agent baselines in three real-world forecasting cases. The code is available at https://github.com/18e0-x/EvoCast.

cs.AI↗
arXiv2026

Continual Learning with Elastic Regularization and Synthetic Replay for Federated MLLM Fine-Tuning

Federated fine-tuning of Multimodal Large Language Models (MLLMs) across distributed networks enables privacy-sensitive adaptation to evolving data streams, yet a fundamental obstacle prevents robust deployment in dynamic environments: catastrophic forgetting, wherein sequential task updates erase previously acquired knowledge across visual, linguistic, and cross-modal representations. Addressing this challenge is especially critical for autonomous networked AI operating in safety-sensitive domains, such as content moderation, where reliable retention of prior knowledge underpins system integrity. To overcome this, we propose Federated Continual Multimodal Learning (FedCMM), a framework that embeds continual-learning safeguards into the federated optimization loop at three complementary levels. At the parameter level, modality-aware elastic weight consolidation computes separate Fisher information matrices for the vision encoder, language backbone, and cross-modal projector, providing granular, asymmetry-aware protection against modality-specific forgetting. At the data level, each client trains a lightweight local generative replay module to synthesize raw-data-free embedding-level multimodal replay tuples without any raw data sharing. At the aggregation level, Task-similarity-aware gradient aggregation autonomously filters and reweights client updates by gradient cosine similarity, suppressing conflicting directions and stabilizing the global learning trajectory. Extensive experiments on two benchmarks demonstrate that FedCMM consistently outperforms recent baselines on accuracy and backward transfer, confirming that holistic, modality-aware optimization enables robust evolutive adaptation across heterogeneous networked AI deployments.

cs.LG↗
arXiv2026

PFAdapter: Hierarchical LoRA Decomposition for Personalized Federated MLLMs

Agentic AI systems are reshaping communications and networking by deploying autonomous intelligent agents capable of collaborative learning while maintaining data privacy at network edges. Within distributed network environments, Multimodal Large Language Models (MLLMs) serve as cognitive engines for edge devices, yet federated fine-tuning faces substantial challenges in balancing global knowledge aggregation with local adaptation under heterogeneous network conditions. Conventional federated protocols typically rely on uniform parameter aggregation, which conflates domain-invariant features with client-specific nuances, thereby resulting in suboptimal personalization and excessive communication overhead. To address these challenges, we propose PFAdapter, a communication-efficient framework introducing hierarchical LoRA decomposition to explicitly separate adapter parameters into global-shared and local-private components. Query and key projections are assigned to global synchronization for capturing universal multimodal semantics across the network, while value and output projections remain localized for edge-specific adaptation. Additionally, orthogonality regularization based on the Frobenius norm enforces strict separation between these components, preventing redundant feature learning. Selective aggregation protocols synchronize only global-shared components across the federated network, preserving local expertise and reducing communication costs by nearly 50%. Extensive experiments on VQA-RAD, SLAKE, Hateful Memes, and CrisisMMD datasets demonstrate that PFAdapter consistently outperforms state-of-the-art baselines, achieving accuracy improvements ranging from 2.4% to 4.8% across diverse edge intelligence tasks. Consequently, our framework establishes an efficient solution for agentic AI deployment in resource-constrained communication networks.

cs.LG↗
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, but how VLM representations differ from vision-only encoders after policy learning, and whether such differences matter for planning, remains unclear. Under a unified VLM-hidden + diffusion-policy paradigm, we compare multiple VLM families/scales (InternVL3 and Qwen3VL) with standard vision-only encoders (ResNet, ViT, and EVA-CLIP) while keeping the downstream planner fixed. We study representation, behavior, and system design. CKA/CCA and Shared--Unique SAE show that policy learning enlarges a common decision subspace, but both branches retain non-transferable residual factors. We further replicate this shared-plus-unique representation pattern on nuPlan using the AsyncDrive planning stack. Latent-intervention policies and scenario-level analysis further show that these residuals are behaviorally meaningful: vision-only encoders are stronger in simple geometry-dominant scenes, whereas VLMs are more effective in semantically complex and interaction-heavy long-tail cases. The two branches also exhibit distinct progress--braking and path-choice tendencies, and an oracle best-of-two VLM+ViT selector reaches 93.58 PDMS on NAVSIM. We convert this complementarity into two lightweight systems: HybridDriveVLA, which selects from a compact cross-model candidate set using a learned trajectory scorer and improves PDMS from 90.80 to 92.10, and DualDriveVLA, a fast--slow variant that invokes the VLM in only 15% of scenarios, achieving 91.00 PDMS with about $1.9\times$ lower latency than the VLM baseline. Code will be available at https://github.com/WilliamXuanYu/HybridDriveVLA.

cs.RO↗
arXiv2026

Variational Quantum Homotopy Perturbation Method to Solve Nonlinear Partial Differential Equations

Solving nonlinear partial differential equations is important in various scientific and engineering applications. Recently, quantum computing was introduced as an alternative computational paradigm for solving nonlinear partial differential equations. In this paper, a new method called the variational quantum homotopy perturbation method is proposed to improve the scalability of solving nonlinear partial differential equations through two aspects. First, the dimension of the Hilbert space remains the same after the nonlinear partial differential equation is linearized through the homotopy perturbation. Second, the solutions are obtained with a variational quantum simulation framework, where the number of qubits is decreased with functional encoding and the depth of parametrized circuits is reduced. The additional contribution of this paper is the introduction of new criteria for selecting the homotopy series truncation order and circuit depth for cost-effective variational quantum homotopy perturbation method. The proposed approach is demonstrated with several examples, including the vorticity transport equation and the reduced magnetohydrodynamics equations.

quant-ph↗
arXiv2026

The quadratic Brown--Erdős--Sós problem for 3-uniform hypergraphs with 8 and 9 edges

The famous and actively studied problem of Brown--Erdős--Sós from 1973 asks for $f^{(r)}(n;s,k)$, the maximum number of edges in an $r$-graph with $n$ vertices in which no $s$ vertices span $k$ or more edges. In this paper, we concentrate on the case $r=3$ and $s=k+2$, with $k\ge2$ fixed and $n\to\infty$; then it is easy to show that the extremal function grows quadratically in $n$. Delcourt and Postle proved that the limit $π(k):=\lim_{n\to\infty} f^{(3)}(n;k+2,k)/n^2$ exists for every $k$. While Brown, Erdős and Sós observed that $π(2)=1/6$ already in the 1970s, the value of $π(k)$ for $3\le k\le 7$ was determined only recently (by various subgroups of Glock, Joos, Kim, Kühn, Lichev, Pikhurko, and Sun). Very recently, Chao, Huang and Liu determined $π(k)$ for every odd $k$. Independently of the last result, we show that $π(9)=1/5$. Also, we prove that $π(8)\le {5053}/{26544}$, which is within $0.0029$ of the best known lower bound $π(8)\ge 3/16$. The new upper bounds are obtained by expressing some previous arguments as a linear program and then using a computer to generate and solve its instances. Our proof of the lower bound on $π(9)$ is based on a finite field construction combined with existing packing results.

math.CO↗
arXiv2026

DuoMatching: Joint-Marginal Distribution Matching for Few-Step Video Generation

Streaming video generation has benefited from distribution matching distillation (DMD), which matches the joint distribution of video frames to a video teacher's approximation of the real video distribution. Although this joint matching mitigates drift during autoregressive rollouts, limitations remain in visual quality and semantic alignment. To address these limitations, we propose DuoMatching, a distribution matching framework that approximates the real video distribution through a unified joint-marginal formulation. On top of existing joint matching formulations, the additional marginal matching objective provides dedicated frame-level supervision from an image generator, transferring complementary visual and semantic priors from it. To apply this frame-level supervision in video generation, we introduce LatentBridge to resolve the latent representation mismatch between the video student and the image teacher. Latent Variation Sampling further distributes such frame-level supervision across distinct temporal segments, reducing redundancy. Experiments demonstrate that DuoMatching improves visual quality, composition, and semantic alignment while largely preserving motion dynamics. Human evaluations show overall preference rates above 80% against all evaluated baselines. The project page is available at https://johnzhan2023.github.io/DuoMatching/.

cs.CV↗
arXiv2026

CtrlWAM: Controllable World Action Models with Aligned Intent and Foresight

World action models (WAMs) jointly predict actions (intent) and visual future (foresight). Standard training adds noise to recorded actions and video simultaneously, but such training paradigms introduce a mismatch: perturbed actions imply counterfactual future visual, while the noised video remains tied to the GT recording. In low-noise regime, the scene geometry and even the dynamic behavior remain clearly visible from the noisy future frames despite the added noise. We present CtrlWAM, which executes perturbed actions in a simulator and pairs them with their noised visual consequences for joint WAM learning. To accommodate the different denoising requirements of video and actions, we introduce warped video--action noise schedules that aim to keep visual layout responsive as action predictions evolve. We further extend the action interface from ego-only control to a variable number of agent streams, allowing a unified model to represent predicted or commanded futures for multiple agents. Driving experiments show more accurate action forecasts, closer agreement between generated video and actions, and better following of supplied commands; robotics experiments show stronger motion fidelity and controllability. Matched controls support the benefit of off-path renders for command following and manipulation fidelity. Together, these findings contribute to a more controllable world action model. Project page: https://ctrl-wam.github.io/

cs.CV↗
arXiv2026

Can Agents Trust Their Skills? Uncovering Unsafe Chains of Trust in Skill-Based LLM Agents

LLM agents increasingly rely on installable skills, which are packages of instructions, code, and resources that equip them with task-specific capabilities and, once installed, can be automatically invoked across subsequent user tasks. This creates a chain of trust in which users delegate authority to agents, while agent frameworks admit skill-provided content into the agents' context with insufficient validation, allowing malicious skills to influence agent behavior under that delegated authority. Yet, little is known about whether this trust model adequately constrains untrusted skill content before it reaches security-sensitive operations, or how frequently such trust violations arise in real-world agents. We present TrustProbe, a framework for uncovering unsafe chains of trust in skill-based LLM agents. First, TrustProbe analyzes agent source code to identify source-to-sink call paths from skill-controlled inputs to security-sensitive operations. Second, it generates semantically realistic SKILL.md seeds with injected canaries and evolves them through feedback-guided scheduling and mutation. Finally, it validates vulnerabilities using an oracle that confirms attacker-controlled flows and verifies observable harm. Across 11 open-source agents, eight with more than 10,000 GitHub stars, TrustProbe identifies 104 taint-style vulnerabilities. Validation on a large corpus of real-world skills collected from public hubs such as ClawHub further shows that 25.1% of skill-agent trials exercise the identified vulnerable paths, with payload injection successfully weaponizing 15 of the vulnerabilities. These results reveal a systematic trust failure in skill-based LLM agents: untrusted skill content can reach security-sensitive operations and exercise authority delegated by users to their agents.

cs.CR↗