Searcharxiv⌕ Search

arXiv subjects

Xiang Zhou

Publications and source records attributed to Xiang Zhou.

At least 19 recordsLinked to original sources

SkillEarth: Teaching General-Purpose Agents to Act on Earth via Skill Evolution

Recently developed remote sensing (RS) agents execute real-world tasks with complex workflows through multi-step tool invocation. These agents, however, typically rely on predefined expertise or parameter updates. They thus require updating expert knowledge or retraining models whenever new tools, tasks, or execution patterns emerge, limiting their continual adaptability in dynamic Earth environments. In this work, we seek to equip general-purpose agents with RS tool-call knowledge that can be automatically extracted from demonstrated trajectories without expert intervention or parameter updates. Specifically, we propose SkillEarth, a self-evolving skill extraction method that externalizes RS tool-call knowledge as reusable, retrievable, and evolvable skill files and refines them through observation-grounded evolution. Unlike existing automatic skill extraction methods that rely on outcome-based reflection to refine skills holistically, SkillEarth grounds skill evolution in real tool-chain execution and combines dual-judge feedback with a grounded observation routing graph for precise component-level refinement. The resulting skills are generalizable, reliable, and transferable: they improve overall RS tool-call performance across three frontier LLMs and four open-source LLMs, are refined through real tool execution feedback, and support cross-model reuse without retraining. Experiments on OpenEarth-Bench and Earth-Bench demonstrate that SkillEarth achieves better tool selection, argument correctness, execution-order consistency, and task completion than existing RS agents and skill extraction methods.

cs.AI↗

VisHead-Spec: Head-Routed Visual Attention for Multimodal Speculative Decoding

Speculative decoding accelerates vision-language models(VLMs) while preserving the target model's outputs, but its speedup diminishes under high concurrency serving as the system becomes more compute bounded as the batch size increases. This problem is pronounced for multimodal speculative decoding as the long range of the image tokens in the prompt can significantly increase the computational cost during draft token prediction. The block-diffusion drafters (e.g. DFlash, DSpark) can effectively speed up the draft model efficiency by predicting multiple draft tokens in one forward pass, but their speedup for VLMs under high concurrency is still modest (<9%). In order to understand this gap, we examine the information flow between different modalities from target model to draft model and identify the sparsity pattern of visual attention between the target and draft models: only small set of key-value heads in the draft model need direct visual feature attention in the context and only visual features from top level layers in the target model need to be attended in the context. We therefore propose VisHead-Spec, which selects a small, fixed set of key-value (KV) heads to attend to all image features from top level layers in the target model and restricts the remaining heads to text features. On Qwen3-VL-8B and MMSpec, VisHead-Spec-8 lets 8 of the drafter's 40 KV heads read the image while keeping accepted length within 0.3% of the block-diffusion draft model baseline (i.e. DSpark). At batch size 64, its throughput gain over autoregressive decoding increases to 16.0%, compared with 9.5% for DSpark; at batch size 1, its speedup remains comparable to DSpark.

cs.LG↗

KVUse: Reusable-KV User Memory for Efficient LLM Recommenders

Memory-augmented recommenders built on large language models (LLMs) maintain a natural-language profile of each user. These profiles are updated by a user profile update path that revises the profile as new interactions arrive and are consumed by downstream task paths like ranking. Because these user profiles are stored as text in the user memory, every ranking LLM call that consumes them must splice them in the prompt and calculate the KV cache of them during the prefill phase. Given these ranking LLM calls are prefill heavy (long prompt but short decoded output) and repeatedly called, these repeated prefill of user profile text dominates the serving. Prefix caching cannot remove it, because these user profiles are preceded by different text in different calls. We present KVUse, which stores the user profiles as reusable key-value (KV) caches instead of text. KVUse trains lightweight LoRA writers by self-distillation, so that the user profile update call produces the reusable KV of the user profile while decoding it, and the LLM can read these reusable KV at new positions and behind new context without recomputation. KVUse also has a customized KV scheduler that uploads the stored reusable KV of the next batch while the current batch is computed, hiding the KV transfer from host memory behind GPU computation. On four InstructRec datasets with Llama-3.1-8B-Instruct, KVUse shows no statistically significant difference in ranking quality from full prefill over continuous memory update rounds. Under saturated load, KVUse raises the throughput of the downstream scoring path by 4.0x-4.5x and of the profile update path by 1.4x-1.9x.

cs.LG↗

Physics-based simulation unlocks scaling laws for generalizable MRI reconstruction

Deep-learning-based magnetic resonance imaging (MRI) reconstruction holds great promise for accelerating acquisitions, yet its clinical translation remains bottlenecked by poor generalizability across unseen scanners, populations, anatomies, and protocols. Gathering sufficiently diverse clinical datasets to overcome this bottleneck is hindered by high acquisition costs, privacy restrictions, and institutional biases. Here, we introduce Scalable Inverse Model (SIM), a framework that systematically co-designs physics-simulated training data, model architecture, and learning paradigms around scalability--leveraging compute scaling to overcome historical data bottlenecks and achieve robust zero-shot MRI reconstruction. Across 14 public and in-house datasets--spanning seven international centers, five vendors, field strengths of 0.3-7 T, over ten anatomical regions, more than ten acquisition protocols, downstream quantitative and molecular tasks, and a 150-scan blinded radiologist study--SIM demonstrates robust zero-shot generalizability, recovering fine anatomical details while suppressing aliasing artifacts where conventional deep-learning methods fail. Attention-map analyses confirm that SIM learns physically explainable, dynamically adaptive reconstruction primitives rather than anatomy-specific priors. SIM establishes a scalable, privacy-preserving, and democratized infrastructure for universal MRI reconstruction, providing a blueprint for generalizable AI across inverse problems in medical imaging.

eess.IV↗

Rethinking One Gene, One Token: Transcriptome-Aware Tokenization for Single-Cell Foundation Models

Single-cell foundation models typically treat each gene as a token, which is both costly and wasteful. A cell can express thousands of genes, so one token per gene makes self-attention expensive, yet many of these genes are detected at low levels and carry too little signal to deserve a token of their own. Genes that are expressed together could instead share a token, leaving separate tokens for genes that stand out. We introduce scTAT (Transcriptome-Aware Tokenization), which pairs a gene-set vocabulary with a per-cell tokenizer. The vocabulary holds overlapping gene sets that recur with stable relative expression across the pretraining corpus. Using these sets raises two difficulties. First, which genes should share a token depends on the cell, whereas existing multi-gene tokens fix the grouping in advance. Second, gene sets of different sizes must be compared with single genes on a common scale. The tokenizer therefore decides per cell, letting gene sets and single genes compete for token positions under an expression score normalized by the square root of set size. Training still reconstructs individual genes, so predictions stay at gene resolution. In frozen-embedding cell-type annotation, scTAT matches the strongest one-gene-per-token model with half the token budget. Its advantage over one-gene-per-token models at the same budget grows as the budget shrinks and the tokenizer assigns more positions to gene sets. Encoding the same genes with gene-set tokens also nearly halves the Transformer's matrix-multiplication cost. These results show that a token need not hold exactly one gene: when the tokenizer decides per cell which genes share a token, a fixed budget covers more of each transcriptome at no cost in annotation accuracy.

q-bio.GN↗

Knowledge and Reasoning Augmented Learning for Language Model Assisted Antimicrobial Decision Support

Background: Antimicrobial decision support requires both clinical knowledge and patient-specific reasoning. We developed Knowledge and Reasoning Augmented Learning (KRAL) to improve language model performance on antimicrobial decision-making tasks. Methods: In this retrospective technical study, a 27-billion-parameter student model underwent guideline-grounded supervised fine-tuning followed by Multi-turn Agentic GRPO using clinical reasoning trajectories. Six configurations were evaluated on 123 clinician-validated clinical antimicrobial treatment cases from Peking Union Medical College Hospital (PUMCH) and 100 JAMA Clinical Challenge cases. A language model judge scored responses, and paired tests compared configurations. Component ablations assessed benchmark performance and training stability. Results: On the PUMCH benchmark, KRAL achieved the highest composite score (0.895 on a 0-1 scale), outperforming all five controls (all P<0.001). On JAMA, KRAL scored 0.600 and exceeded the base model with and without retrieval (P = 0.0048 and 0.0499, respectively). Reinforcement learning ablation experiments demonstrate that the reward smoothing can significantly improve the performance and training stability of vanilla GRPO. Conclusions: KRAL improved performance on a single-center clinical antimicrobial benchmark and publicly available JAMA clinical challenge cases by combining guideline-grounded learning with Multi-turn Agentic GRPO. Further external clinical validation is needed to evaluate the practical utility and effectiveness of KRAL.

cs.AI↗

Adaptive Aerial Loco-Manipulation Policy Learning via LLM-Assisted Proactive Task Generation

Loco-manipulation extends the workspace of conventional manipulation by combining mobility with physical interaction, while aerial loco-manipulation further expands this capability into three-dimensional space, enabling robots to reach environments inaccessible to ground-based systems. Moving to an aerial platform, however, gives rise to a demanding control problem, since the vehicle must track trajectories accurately in three dimensions while the motion of the arm and its interaction with the environment continuously shift the center of mass, alter the inertia, and impose force and moment disturbances on the body. Conventional controllers address these effects through explicit models and estimation, whereas learning-based controllers infer dynamics from experience, yet control rewards provide no direct target for the response information retained from history, and fixed task sampling can leave persistent tracking difficulties insufficiently explored. We present ALMA, an aerial loco-manipulation framework for adaptive flight control under external arm commands. ALMA employs a contextual flight policy that adapts from recent measurements and actions, using action-conditioned prediction of future measurement representations to train its context to retain response information relevant to flight control. To expand the experience available for this adaptation, an evolutionary process uses a large language model (LLM) to analyze observed control difficulties and propose new structures of arm motion for subsequent training. In simulation, ALMA reduces position root-mean-square error (RMSE) by up to 24.7% relative to the strongest baseline, while the same policy transfers zero-shot to the aerial manipulator and reduces hover position RMSE by 79.6% during periodic arm motion relative to the real-world baseline.

cs.RO↗

Probabilistic Energy Modeling Unifies Versatile Swarm Control

Robotic swarms must coordinate across diverse missions, yet existing controllers often require redesign, retuning, or retraining when objectives change. Here we show that probabilistic energy modeling provides a unifying representation for versatile swarm control, in which diverse missions emerge from compositions of shared interaction principles rather than task-specific control formulations. By exploiting the correspondence between Gibbs random field energy minimization and structured reward maximization, we encode fundamental swarm behaviors as reusable energy terms within a unified probabilistic objective. Decentralized policies trained within this representation are composed into a single controller through lightweight policy gating, without retraining. An interpretable action-attention mechanism captures non-uniform local interactions, while dimensionless normalization supports transfer across swarm scales and platforms. Demonstrated on aerial swarms, the unified policy generalizes zero-shot across six mission classes, 10-100 agents, two hardware platforms, and indoor cluttered and GPS-denied woodland environments. Large-scale simulations achieve 0.96-1.00 task success rates with 1-10 ms per-agent inference latency, while physical deployments confirm zero-shot transfer to real platforms and unstructured environments. These results establish probabilistic energy modeling as a scalable and transferable framework for decentralized swarm control.

cs.RO↗

$H^2(Ω)$ A Priori Error Bounds for Boundary-Adapted PINNs Solving Elliptic Dirichlet Problems

We study the numerical resolution of uniformly elliptic Dirichlet boundary value problems (BVPs) by minimization of the PDE residual over a hypothesis space of Neural Networks made to satisfy exactly the boundary conditions through multiplication by a user-defined \emph{distance-to-boundary approximation} $ρ$. While an $L^2(\partialΩ)$ boundary penalty cannot generally guarantee $H^2(Ω)$ error control, we show that exact enforcement can, provided $ρ$ is chosen appropriately. More precisely, we identify a sharp sufficient condition on $ρ$ preserving the regularity needed for $H^2(Ω)$ error control, provide practical constructions of such $ρ$ for general smooth domains $Ω$, and refer to the resulting networks as \emph{boundary-adapted}. For boundary-adapted Rectified Quadratic Unit (ReQU) and hyperbolic tangent (tanh) networks, we then derive explicit high-probability $H^2(Ω)$ convergence rates by combining approximation-theoretic and statistical-learning arguments, making explicit the dependence on $ρ$. Numerical experiments support our theoretical results, showing that appropriate choices of $ρ$ improve accuracy and convergence, while poorly chosen distance functions can substantially degrade the solution. Our proof also yields new VC-dimension bounds for hypothesis spaces of higher-order derivatives of ReQU and tanh networks, together with new approximation bounds for shallow ReQU networks in higher-order Sobolev norms, all of which are of important independent interest.

math.NA↗

UEPS: Robust and Efficient MRI Reconstruction

Deep unrolled models (DUM) have become the state-of-the-art for accelerated MRI reconstruction, yet their robustness under domain shift remains a critical barrier to clinical adoption. In this work, we identify coil sensitivity map (CSM) estimation as the primary bottleneck limiting generalization. To address this, we propose UEPS, a novel DUM architecture featuring three key designs: (i) an Unrolled Expanded (UE) design that eliminates CSM dependency by reconstructing each coil independently; (ii) progressive resolution, which leverages k-space-to-image mapping for efficient coarse-to-fine refinement; and (iii) sparse attention tailored to MRI's 1D undersampling nature. These physics-grounded designs enable gains in robustness and computational efficiency. We construct a large-scale zero-shot transfer benchmark comprising 10 out-of-distribution (OOD) test sets spanning diverse clinical shifts, anatomy, view, contrast, vendor, field strength, and coil configurations. Extensive experiments demonstrate that UEPS consistently and substantially outperforms existing DUM, end-to-end, diffusion, and untrained methods across all OOD tests, achieving state-of-the-art robustness with low-latency inference suitable for real-time deployment.

eess.IV↗

PA-LOCO: Learning Perturbation-Adaptive Locomotion for Quadruped Robots

Numerous locomotion controllers have been designed based on Reinforcement Learning (RL) to facilitate blind quadrupedal locomotion traversing challenging terrains. Nevertheless, locomotion control is still a challenging task for quadruped robots traversing diverse terrains amidst unforeseen disturbances. Recently, privileged learning has been employed to learn reliable and robust quadrupedal locomotion over various terrains based on a teacher-student architecture. However, its one-encoder structure is not adequate in addressing external force perturbations. The student policy would experience inevitable performance degradation due to the feature embedding discrepancy between the feature encoder of the teacher policy and the one of the student policy. Hence, this paper presents a privileged learning framework with multiple feature encoders and a residual policy network for robust and reliable quadruped locomotion subject to various external perturbations. The multi-encoder structure can decouple latent features from different privileged information, ultimately leading to enhanced performance of the learned policy in terms of robustness, stability, and reliability. The efficiency of the proposed feature encoding module is analyzed in depth using extensive simulation data. The introduction of the residual policy network helps mitigate the performance degradation experienced by the student policy that attempts to clone the behaviors of a teacher policy. The proposed framework is evaluated on a Unitree GO1 robot, showcasing its performance enhancement over the state-of-the-art privileged learning algorithm through extensive experiments conducted on diverse terrains. Ablation studies are conducted to illustrate the efficiency of the residual policy network.

cs.RO↗

Curvature bounds for the cost of coherence in conditional optimal transport

We study the gap between the integrated direction-wise kinetic minima and the minimum expected Dirichlet energy of a joint particle realization of a prescribed two-parameter family of conditional probability measures. Under regular coarea assumptions with compact connected fibres and a star-shaped parameter domain, we derive a curvature lower bound valid for every admissible law of Sobolev maps realizing the prescribed smooth positive conditional measures. Our method combines the weighted-Poisson characterization of local optimal-transport velocities with an exact energy-residual identity, a weak Stokes formula, and explicit transport constructions. The minimum excess vanishes exactly when the Lie-bracket curvature vanishes throughout the parameter domain, whereas at a nonzero-curvature centre it is of order $r^4$ as $r\to0$ on squares of half-side length $r$ with unnormalized area measure. In a two-phase torus model fitted to electricity-demand data, numerical tests illustrate this quartic order for explicit couplings, with radial transport approximately halving the ordered construction's excess on the tested small squares.

math.OC↗

Sage: Formalization with Semantic Correction

While neural theorem provers have achieved impressive milestones in formal mathematics, they largely operate on the assumption that faithful Lean 4 formal statements are already provided. Translating informal natural language into a formal language is a critical data bottleneck plagued by an "illusion of rigor": standard type-checkers accept statements that compile but drop hypotheses, introduce vacuous truths, or subtly alter mathematical bounds. To resolve this, we introduce Sage (Semantic Agent-Guided Formalization Engine), an agentic framework that replaces monolithic translation with a four-stage decomposed generation pipeline coupled with a dual-signal semantic correction loop. By pairing Lean 4 compiler diagnostics with multi-dimensional semantic feedback, our correction loop enforces mathematical fidelity alongside syntactic validity. By explicitly accounting for the gap between open-ended queries and declarative formal targets, our pipeline prevents models from achieving high formalization rates by guessing unverified answers (exhibiting a 70.9% answer leakage rate in monolithic baselines). Consequently, Sage suppresses leakage to 2.7% while achieving 73.3% pass@4 joint compilation and semantic fidelity on the Omni-MATH without proofs (compared to 42.0% for a fine-tuned Goedel-Formalizer-V2 baseline). Finally, on IMO-Unformalized, a novel frontier of 175 unformalized International Mathematical Olympiad problems, Sage demonstrates effective zero-shot generalization with 87.4% pass@4 verified fidelity compared to just 19.4% for the baseline, winning over 79% of blind pairwise evaluations.

cs.LG↗

Geometry-adaptive Ambisonic encoding for sparse microphone arrays of variable topology using physics-informed diffusion

Ambisonics delivers compact scene based spatial audio representation, yet higher order Ambisonic encoding poses difficulties for wearables and embedded hardware. Their microphone arrays are often sparse, irregular, and constrained by device specific boundary conditions. These factors make the spherical-harmonic (SH) domain encoding ill conditioned: inverse filtering amplifies noise, while deterministic neural encoders may overfit to array-specific responses or smooth ambiguous higher-order components. This paper presents DiffM2A, a geometry-adaptive conditional diffusion framework for robust Ambisonic encoding from sparse MAs with variable topologies. Its Geometry-Adaptive Spherical Harmonic Projection (GASHP) front-end constructs boundary-aware SH steering functions and applies an energy-normalized modal projection, mapping array-dependent observations to a common modal representation without explicit pseudo-inverse computation. A dual-branch Elucidated Diffusion Model then estimates complex Ambisonic coefficients, conditioned on both the raw microphone spectra and GASHP features. Sound intensity and rotational equivariance losses further enhance inter-channel phase consistency and structured behavior across SH subspaces. Evaluations on both first- and second-order Ambisonic encoding tasks, using simulated room-acoustics and real-world LOCATA recordings, demonstrate that DiffM2A outperforms conventional and neural baseline methods on signal fidelity, spectral accuracy, spatial coherence, and binaural cue preservation. Additional experiments show that these gains are largely retained across unseen five-microphone layouts and under mismatched open-array and rigid-sphere boundary models.

eess.AS↗

From coherence to mixedness: a driver of barren plateaus in variational quantum algorithms

Variational quantum algorithms (VQAs) are a leading approach for near-term quantum advantage. However, their training is often hindered by barren plateaus (BPs). We present a framework based on observational entropy. The framework separates the coherent part of a quantum state from its incoherent part. We define the coherence fraction $η$ as the ratio of coherent to total contribution. This quantity measures how much of the coherence capacity is in a usable form. Using a 4-qubit Ising model and a hardware-efficient ansatz, we show that $η$ decreases monotonically during optimization, while the coherence capacity remains nearly constant. Our results show that $η$ provides an early indication of gradient collapse. It outperforms conventional diagnostics such as the gradient norm and entanglement entropy. This work offers a resource-theoretic explanation for BPs. It also provides a basis for real-time monitoring of coherence loss on noisy devices.

quant-ph↗

PILOT: Privileged Imitation Learning for End-to-End Motion Planning of Autonomous UAVs under Partial Observability

Autonomous navigation in cluttered environments is hampered by partial observability and dynamic constraints. This paper presents PILOT, a constraint-aware privileged imitation learning framework for vision-based end-to-end UAV motion planning under partial observability. The framework distills planning strategies from a computationally intensive optimal control expert into a student policy regularized toward safety and dynamic requirements via a dual-objective loss function. To mitigate partial observability, a spatiotemporal perception fusion module using a Temporal Convolutional Network (TCN) is developed to integrate historical depth images and odometry. This module infers task-relevant latent context from historical observations, enhancing spatial awareness beyond the instantaneous FOV without maintaining persistent map memory. A trajectory parameterization layer mapping network outputs to a structured trajectory, while enabling explicit continuity, dynamic-consistency, and obstacle soft penalties during training, encouraging constraint satisfaction for unseen observations without formal guarantees. Simulations on quadrotor and fixed-wing aircraft demonstrate that PILOT achieves performance comparable to the privileged expert while reducing computational overhead by over 80\%. Successful indoor and outdoor zero-shot deployment confirms the practical feasibility and cross-domain generalization of the planner.

cs.RO↗

Taming the Loss Landscape of PINNs with Noisy Feynman-Kac Supervision: Operator Preconditioning and Non-Asymptotic Error Bounds

Physics-Informed Neural Networks (PINNs) often train slowly or fail to converge on challenging partial differential equations (PDEs), a behavior recently linked to severely ill-conditioned loss landscapes inherited from the underlying differential operator. We study PINNs augmented with a pointwise data-fidelity term, added at a few points in the domain to the standard residual and boundary losses. We show that this supervision term acts as an operator-level preconditioner: for suitable weights, our comparison bounds guarantee a substantially smaller condition number than under the standard PINN loss, independently of how the pointwise labels are obtained. For a broad class of PDEs admitting a Feynman-Kac (FK) representation, we generate such labels by Monte Carlo averages of the FK functional, resulting in what we call ``FK-PINNs", and using the excess risk decomposition approach, we derive non-asymptotic $L^2(Ω)$-error bounds for FK-PINNs with $\tanh$ activation trained by finitely many steps of gradient descent. Along the way, we establish pseudo-dimension bounds for first- and second-order derivatives of $\tanh$ neural networks, which are of independent interest and, to the best of our knowledge, new. Numerical experiments on Poisson, Schrödinger, mean exit time, and committor problems corroborate the theory, and show that FK-PINNs can successfully solve PDEs for which standard PINNs exhibit severe failure modes.

stat.ML↗

TCellAlign: Cross-study T-cell Populations Alignment with Nomenclature-Guided Multi-Agent Workflow

Cell type standardization plays a central role in integrating biological knowledge across single-cell studies. While standardized resources (e.g., Cell Ontology, Nomenclature Frameworks) provide unified vocabularies of cell populations, scientific publications and public datasets continue to use heterogeneous study-specific labels, making cross-study comparison difficult even when biologically equivalent cell populations are described. In this work, we are the first to formulate this challenge as an evidence-grounded cell population alignment problem and propose TCellAlign, a multi-agent framework that includes literature retrieval, information extraction, nomenclature-guided label alignment, and evidence-based adjudication. This modular design preserves the original terminology and supporting evidence reported by each study while producing standardized labels that can be compared across studies. We further construct a manually validated benchmark dataset linking study-specific labels, CZ CELLxGENE annotations, and standardized T-cell nomenclature across 44 manually curated, published studies (including over seven million cells) spanning four biological categories: healthy, cancer, infectious disease and inflammatory diseases. Across the evaluated tasks, TCellAlign achieves stronger semantic agreement than ontology-based baselines and maintains transcriptomic coherence with both open-source and closed-source large language models (LLM) backbones. By connecting literature, datasets, and expert's nomenclature, TCellAlign enables consistent interpretation of T-cell subtypes and states across studies, facilitating biological knowledge integration and the development of future foundation models built upon standardized cellular representations.

q-bio.QM↗