Search arXiv⌕ Search

arXiv subjects

Jingbo Wang

Publications and source records attributed to Jingbo Wang.

At least 19 recordsLinked to original sources
arXiv2026

Q-MERGE: Parallelising Quantum State Preparation for Large-Scale Classical Data

Quantum processing of classical data fundamentally relies on efficiently mapping classical datasets onto the amplitudes of quantum states. Preparing large amplitude-encoded states, however, remains a major bottleneck in quantum computing. In this paper, we introduce a novel state-preparation framework we call Q-MERGE, that addresses this bottleneck by partitioning a large target state into $M$ $n$-qubit segments, preparing these segments independently and in parallel, and then coherently combining them into a single amplitude-encoded state using SELECT-SWAP operations and measurement. Q-MERGE is agnostic to the segment-level preparation method, allowing existing techniques to be applied to smaller subproblems while providing a tunable trade-off between circuit depth and qubit count. With mid-circuit measurement and preparation-register reuse, the required ancilla qubits can be reduced from O(Mn) to O(n). Applied to a real-world $32 x 1024$ ultrasound dataset, Q-MERGE achieves an infidelity of $3.5 x 10^{-9}$, compared with $6.2 x 10^{-3}$ for direct preparation using the same underlying method, a more than six orders of magnitude improvement at equal CNOT circuit depth. We demonstrate its feasibility experimentally with a 53 qubit circuit comprised of 954 1-qubit gates and 478 2-qubit gates, on the Quantinuum System Model H2 trapped-ion quantum computer. The prepared state is validated using shadow-overlap tomography. Numerical analysis of Haar-random states indicates consistent success probability for up to $M=10^7$ segments, supporting the scalability of Q-MERGE for encoding massive classical datasets.

quant-ph↗
arXiv2026

Latent Visual Contrastive Reasoning with Multi-Branch Exploration

Multimodal large language models have made substantial progress in visual question answering, but complex visual reasoning often requires locating and comparing multiple pieces of question-relevant visual evidence. Existing methods that move intermediate reasoning into the visual latent space usually follow a single linear reasoning trajectory, which can be limiting when multiple visual hypotheses or competing evidence regions are relevant. We propose Latent Visual Contrastive Reasoning, a branch-aware latent reasoning framework that generates multiple question-conditioned latent visual branches as soft hypotheses over image tokens. To train these branches, our method uses bounding-box evidence annotations to build target evidence representations and introduces a soft contrastive branch-evidence alignment objective. This objective encourages evidence-relevant branches to align with annotated visual regions while contrasting them against negative evidence. We instantiate our method on Qwen2.5-VL backbones and train it with standard supervised fine-tuning. Experiments on perception-intensive visual reasoning benchmarks show that the proposed method achieves overall improvements over the corresponding backbone and single-path latent reasoning baselines at both 3B and 7B scales.

cs.CV↗
arXiv2026

Designing Quantum Solutions to the Closest String Problem with the Non-Variational QWOA

Quantum computing offers promising approaches to combinatorial optimization problems whose search spaces grow exponentially with problem size. The Closest String problem belongs to this category, which seeks a string that minimizes the maximum Hamming distance to any string in a given collection of reference strings. In this work, we present an end-to-end quantum implementation of the Closest String problem using the Non-Variational Quantum Walk Optimization Algorithm (NV-QWOA), evaluating its resource requirements and solution amplification capabilities. Our automated workflow converts arbitrary problem instances into quantum circuits expressed in one- or two-qubit gates that encode candidate solutions, evaluate the objective function, and apply NV-QWOA optimisation to increase the measurement probabilities of high-quality solutions. Analytical resource estimates yield a qubit complexity of $\mathcal{O}(nm\log k)$ and a circuit depth of $\mathcal{O}(pm\log^2 n + pn\log n)$, where $n$ is the string length, $m$ is the number of reference strings, $k$ is the alphabet size, and $p$ is the number of NV-QWOA iterations. We introduce a revised prescription for the phase-separation parameter, $γ$, that depends on problem size but requires no instance-specific optimization. This prescription mitigates excessive phase wrapping modulo $2π$, which can weaken the relationship between objective values and the phases guiding the optimization dynamics. Across all simulated problem sizes, the algorithm consistently amplifies optimal and near-optimal candidates. This work provides an explicit gate-level implementation, an improved phase-scaling method, and a foundation for hardware experiments and benchmarking against classical algorithms.

quant-ph↗
arXiv2026

ReactiveBFM: Reactive Closed-Loop Motion Planning Towards Universal Humanoid Whole-Body Control

While current Behavior Foundation Models (BFMs) provide robust control priors for humanoids, they only execute pre-defined reference motions. As a result, they are vulnerable to environmental shifts and incapable of reactive whole-body coordination. Naively cascading them with generative motion planners fails to achieve true reactivity, as inevitable tracking discrepancies induce fatal cumulative exposure bias. To bridge this gap, we propose ReactiveBFM, a real-time closed-loop planning-control framework. At its core, we effectively mitigate exposure bias via a scheduled prefix sampling curriculum, forcing the generative planner to actively learn error-recovery behaviors from imperfect physical states rather than ground-truth trajectories. Systematically, to reconcile the severe latency mismatch between auto-regressive planning and high-frequency tracking, we introduce an asynchronous replanning mechanism. Combined with trajectory chunking to temporally ensemble spatial references, our system guarantees spatio-temporally fluid execution without physical jitter. Deployed on the Unitree G1 humanoid, ReactiveBFM demonstrates unprecedented physical agility across a vast repertoire of text-conditioned closed-loop motions. Notably, ReactiveBFM achieves zero-shot moving target reaching, showcasing intricate whole-body coordination and on-the-fly replanning. In sim-to-sim benchmarking under severe perturbations, ReactiveBFM achieves a 93.1% success rate, significantly outperforming cascaded open-loop baselines by 28.6%.

cs.RO↗
arXiv2026

SCDM: Scalable Causal Discovery in Nonlinear Temporal Systems with Meta-Learning

Causal discovery from nonlinear multivariate time series is challenging in high-dimensional systems, where the number of candidate directed relations grows quadratically with the number of variables. Existing methods are often limited by target-wise model fitting, repeated conditional testing, or expensive graph-search procedures. We propose SCDM, a shared-parameter meta-learning framework for scalable temporal causal discovery. SCDM treats each target variable as a task while learning a shared temporal predictor and a shared causal-strength matrix across tasks. Each meta-episode updates only a subset of target-variable tasks, allowing the model to accumulate structural evidence across the full system without exhaustive target-wise optimization. After training, a post-training graph readout converts the learned predictor into continuous directed causal scores for threshold-independent evaluation. We also provide a finite-error recovery principle showing that, under fully observed delayed-system assumptions and a positive risk-gap condition, thresholding SCDM scores recovers the graph when statistical, meta-optimization, and readout errors are below the population margin. Experiments on controlled synthetic, neuroimaging-inspired, high-dimensional, and real-world temporal benchmarks demonstrate competitive causal recovery and improved scalability in large temporal systems.

cs.LG↗
arXiv2026

UniWAM: Unified World-Action Model

Vision-language-action models benefit from the understanding and reasoning capabilities of pretrained vision-language models, but action-only supervision provides limited grounding in world dynamics. Conversely, world-action models inherit spatiotemporal priors from video generation models, yet remain limited in semantic understanding and reasoning under distribution shifts. We introduce UniWAM, a unified architecture that integrates a physical reasoner, a world generator, and an action predictor to jointly learn semantic understanding of the physical world, visual generation, and action prediction. To ensure the quality of the training data, we developed a rigorous data cleaning and annotation pipeline for both human egocentric data and robot data. To adapt the vision-language component to embodied tasks while preserving its inherited language capabilities, we represent low-level actions in natural language and introduce a pre-training recipe that assigns complementary supervision from visual question answering (VQA) data, human egocentric data, and robot demonstrations to the appropriate model components. During post-training, future visual noise augmentation reduces reliance on precise future predictions, while history-conditioned flow matching uses encoded action history to initialize action generation. Together, these designs significantly reduce denoising steps while maintaining performance. UniWAM achieves state-of-the-art (SOTA) performance across multiple evaluations, including in-distribution performance, robustness, generalization, instruction following, and long-horizon task execution. Furthermore, we uncover a log-linear scaling law of unified human-robot co-training, demonstrating the effectiveness of large-scale pre-training on a mixture of human and robot data.

cs.RO↗
arXiv2026

Taking the PULSE of Medical Reasoning: Step-Level Retrieval Timing for Knowledge-Grounded Inference

Reliable medical reasoning requires LLMs to dynamically access external biomedical evidence to improve factual trustworthiness and reasoning accuracy. However, indiscriminate retrieval can introduce noise and latency, raising a critical retrieval-timing problem: at each reasoning step, the system must decide whether the model's internal knowledge is reliable or external evidence is needed. We propose PUlSE, a lightweight probing method that uses the model's internal representations as intrinsic signals for step-level retrieval control. Our posterior trajectory analysis reveals that erroneous steps exhibit substantially higher inter-layer volatility than correct ones, suggesting unstable internal knowledge. Based on this finding, PUlSE trains a compact probe on inter-layer geometric features to detect factual errors upon step completion, and uses it as an adaptive trigger for retrieval and step regeneration. Across four medical benchmarks, PUlSE-timed retrieval matches exhaustive per-step retrieval in accuracy while reducing retrieval calls by 72%, and can transfer to an unseen legal domain, indicating potential domain-agnostic generality. Overall, our study provides an initial step toward shifting retrieval-trigger design from surface-level confidence estimates to internal activation dynamics for adaptive access.

cs.CL↗
arXiv2026

PingpongDuo: Agile Humanoid Table Tennis Doubles with Safe Team Coordination

Unlike singles, humanoid table tennis doubles requires teammates to strike in strict alternation while sharing a confined workspace. Rapid repositioning can compromise striking accuracy, while uncoordinated teammate movements introduce collision risks. We present PingpongDuo, a hierarchical framework for agile humanoid doubles with safe team coordination. At the team level, a planner jointly selects positioning targets and uses learned risk estimates to filter risky proposals, preferentially preserving the current striker's target. At the individual level, separate footwork and striking skills combine agile movement with precise racket control. The skills are guided by human motion references, including footwork collected from human doubles play, and trained from states produced by one another to support rapid transitions. A racket-arm residual further refines strikes after movement using return feedback from a simulator calibrated with real ball-flight and contact measurements. In simulation, PingpongDuo improves the ball-return success rate by 13.2% over the state-of-the-art baseline, while achieving approximately 2 cm positioning error during clearance. In real-world experiments, a robot completes a 2.15 m clearance-and-recovery cycle in 2.4 s with a peak speed of 2.23 m/s, while two robots sustain more than 10 consecutive strikes in strict alternation.

cs.RO↗
arXiv2026

HumanX: Toward Agile and Generalizable Humanoid Interaction Skills from Human Videos

Enabling humanoid robots to perform agile and adaptive interactions has long been a core challenge in robotics. Current approaches are bottlenecked by either the scarcity of realistic interaction data or the need for meticulous, task-specific reward engineering, which limits their scalability. To narrow this gap, we present HumanX, a unified framework that compiles human video into generalizable, real-world interaction skills for humanoids, without per-task design. HumanX integrates two co-designed components: XGen, an anchor-centric data generation pipeline that yields physically plausible robot interaction data from monocular video and augments it to create a diverse, scalable training dataset; and XMimic, an interaction-prioritized learning framework to achieve generalizable interaction skills from seconds of demonstration. Evaluated across five distinct domains--basketball, football, badminton, cargo pickup, and reactive fighting--HumanX successfully acquires 9 different skills and transfers them zero-shot to a physical Unitree G1 humanoid. The learned capabilities include complex maneuvers such as pump-fake turnaround fadeaway jumpshots without any external perception, as well as interactive tasks like sustained human-robot passing sequences over 10 consecutive cycles--learned from a single video demonstration. Our experiments show that HumanX achieves higher generalization success than prior methods, demonstrating a scalable pathway within a class of anchorable interactions for learning versatile, real-world robot interactive skills.

cs.RO↗
arXiv2026

Discrete Forcing: Infusing Discrete Guidance into Continuous Denoising for Few-Step Action Experts

Efficient action generation in vision-language-action (VLA) models requires capturing both coarse action structure and fine-grained details. Discrete action tokens provide compact structural representations but sacrifice precision, while continuous action tokens offer high precision but often require multiple denoising steps. We introduce Discrete Forcing, a flow-matching framework that combines these representations through an explicit coarse-to-fine generation process. It first predicts discrete action tokens to establish a coarse action structure, then uses them to guide continuous action refinement. The discrete and continuous components share a common diffusion transformer backbone with specialized branches, maintaining a parameter count comparable to a conventional single-branch model while requiring only one forward pass per branch. Extensive evaluations across multiple benchmarks demonstrate improved performance and faster inference over a parameter-matched continuous action expert, with consistent performance gains as model capacity increases. Real-world experiments further demonstrate improvements on high-precision and dynamic manipulation tasks.

cs.RO↗
arXiv2026

Combating Instruction Conflict via Energy-Driven Latent Conflict Detection

Large Language Models (LLMs) are increasingly deployed with hierarchical instructions, yet they remain vulnerable to conflicts in which user directives override system-level constraints. Existing defense mechanisms predominantly focus on static input inspection and therefore fail to detect Response Drift, a phenomenon in which the model's final response violates system-level constraints despite seemingly compliant inputs. To bridge this gap, we introduce ELCD, a response-level latent conflict detector for post-generation, pre-delivery verification. Given the full generated output, ELCD constructs a composite hidden-state representation by concatenating the final-token embedding with the mean-pooled response embedding. It then optimizes a pairwise margin ranking objective to separate compliant and drifting responses in latent space. Extensive experiments across five mainstream LLMs ranging from 1.5B to 14B parameters demonstrate that ELCD significantly outperforms competitive baselines. Notably, it improves the PR-AUC on Llama-2-7B by approximately 30 percentage points and reduces the False Positive Rate at 95% TPR (FPR95) on Mistral-7B to 2.67%. These results suggest that ELCD provides a promising approach for latent instruction-conflict detection in open-weight or self-hosted LLM deployments.

cs.CL↗
arXiv2026

MoSAT: Human Motion Generation from Spatial Audio and Textual Description

Human motion is shaped by both external acoustic events and behavioral intent: spatial audio conveys environmental cues that elicit or guide a response, while text specifies the desired action and how it should be performed. In this paper, we study the novel task of human motion synthesis jointly conditioned on spatial audio and natural language, a problem that has been largely overlooked in previous research. To support this task, We introduce STAM, a dataset of motion sequences paired with spatial audio and detailed textual annotations whose rich vocabulary affords precise and nuanced specification of human motions. We further introduce MoSAT, a latent flow-matching framework for full-body motion generation jointly conditioned on natural-language intent and directional spatial-audio cues through hierarchical cross-attention before generating motion. Such a hierarchical design enhances temporally coherent and semantically aligned motion sequences. We also develop tri-modal evaluators for comprehensive evaluation on this novel task. Extensive experiments show that MoSAT achieves the SOTA performance by leveraging spatial audio's intrinsic motion-shaping properties alongside textual semantics, enabling precise and diverse motion in various scenarios.

cs.GR↗
arXiv2026

Hyperlyve: Hyperplane Partitioning for Neural Lyapunov Verification

This work introduces HyParLyVe (Hyperplane Partitioned Lyapunov Verifier), a novel algorithm for sound and complete verification of neural Lyapunov candidates by interpreting shallow ReLU networks as hyperplane arrangements. This perspective reduces positive definiteness verification to a finite set of vertex evaluations, and the decrease condition to a bounded optimization problem over each region. We formally prove correctness of the proposed verification procedures and demonstrate that HyParLyVe achieves significant speedups over state-of-the-art methods.

eess.SY↗
arXiv2026

PASSAGE: Scaling Scene-Aligned Motion Learning for Perceptive Humanoid Traversal in Cluttered Environments

Humanoid robots can step over, squeeze past, and duck under obstacles, but learning to select and coordinate these behaviors from onboard perception remains challenging. Many existing approaches rely on task-specific reinforcement-learning objectives or curated motion libraries, making broad behavioral coverage costly. We present PASSAGE, a perception-conditioned planner--tracker framework for humanoid traversal. Using virtual reality and inertial motion capture, we collect 100 h of scene-aligned human motion across 1,500 cluttered scenes. A conditional flow-matching planner generates short-horizon references from motion history, a local destination, and a robot-centric multi-layer elevation map, while a perceptive whole-body tracker executes them at 50 Hz with geometric feedback. Real-time chunking promotes inter-chunk consistency, and planner-side RL post-training under the frozen tracker further improves closed-loop performance. Without skill annotations or obstacle-specific policies, one planner--tracker pair selects and composes traversal behaviors across unseen geometries. In simulation, component ablations quantify the contribution of each stage. Across three independent training seeds, scaling captured data from 6 to 100 h increases mean contact-free success from 48.1% to 68.9% on held-out scenes, while the final model with validated scene augmentation reaches 70.3%. The fully onboard system integrates egocentric 3D LiDAR perception, online occupancy mapping, 6.25 Hz planning, and 50 Hz control on a Jetson AGX Orin; tests across 50 unseen physical layouts demonstrate traversal without prebuilt maps or offboard computation.

cs.RO↗
arXiv2026

From Block-encoding to Generalized Quantum Signal Processing: Principles, Algorithms and Applications

Modern quantum algorithms are increasingly formulated as coherent procedures for implementing polynomial transformations of operators and singular values. This perspective provides a powerful and unifying language for quantum algorithm design, connecting a wide range of distinct problems through five closely related key tools: block-encoding, qubitization, QSP, QSVT and GQSP. Block-encoding embeds non-unitary matrices into larger unitaries; qubitization converts block-encodings into structured operators; QSP, QSVT and GQSP enable polynomial transformations with near-optimal query complexity. Together, these techniques form a general toolkit for transforming matrix functions into implementable quantum circuits. This paper develops these techniques from first principles as a unified framework for constructing quantum algorithms. We apply this framework to representative applications to highlight design principles and demonstrate how distinct algorithms can be constructed from a unified sequence of operator transformations. A central contribution is a systematic decision workflow for selecting the appropriate approach according to the operator structure and the desired transformation polynomial. This perspective clarifies when direct GQSP or through qubitization, or Laurent expansion, or QSVT is most appropriate. We organize algorithmic design into an end-to-end pipeline: identifying the target matrix function, constructing an appropriate block-encoding, determining the relevant spectral domain, designing a polynomial or Laurent-polynomial approximation, synthesizing the phase factors, and translating the transformation into an executable quantum circuit. By applying this unified framework to example applications, we showcase a practical methodology for reasoning, designing, and implementing quantum algorithms based on polynomial transformations.

quant-ph↗
arXiv2026

Efficient quantum state preparation on Quantinuum hardware

Preparation and verification of specific quantum states is an important capability for quantum devices to realise advantages over classical computations and algorithms. In this work, we have demonstrated an end-to-end framework that combines resource-efficient quantum state preparation with rapid, robust fidelity verification on near-term quantum hardware. By experimentally preparing and validating a structured complex quantum state encoding a digitized acoustic signal on the Quantinuum H2-1 trapped-ion platform, we achieved a high hardware fidelity of $F_{\mathrm{hw}} = 0.929$. Crucially, this milestone was realized without relying on idealized assumptions or deep fault-tolerant overhead, but rather through resource-minimal circuits optimized for NISQ-era and early fault-tolerant devices. Furthermore, we addressed a key limitation in current quantum state certification. While validation methods like shadow overlap work well for random states, their sample complexity can become prohibitively high for the structured states used in practical algorithms. We mitigate this by introducing a pre-measurement basis-change technique that reduces the verification parameter, $τ$, by over 10 orders of magnitude for structured targets. This approach tightens the theoretical certification guarantees of the shadow overlap method and integrates tensor-network preparation and shadow validation into a unified workflow. These results shift the paradigm of how structured classical data can be mapped to and verified on quantum hardware under realistic noise and measurement budgets. By compressing a robust verification procedure to just 1,000 measurement shots, this framework offers an immediate, scalable benchmarking standard.

quant-ph↗
arXiv2026

CT-$Δ$Bench: A Benchmark for Longitudinal 3D Medical Imaging Difference Reporting with Vision-Language Models

In medical imaging, the clinical value of Computed Tomography (CT) lies not only in depicting current disease status, but crucially in enabling longitudinal comparison of serial scans to determine disease evolution, a process that underpins response assessment, recurrence detection, and ongoing patient management. Yet, despite this central role of temporal comparison in clinical decision-making, existing medical foundation models remain largely confined to single-study understanding, leaving temporally grounded cross-examination insufficiently addressed. To address this gap, we study longitudinal imaging difference reporting, a task in which a model takes two temporally separated scans from the same patient and generates a clinically meaningful report describing interval changes between them. We introduce CT-$Δ$Bench, a dedicated benchmark for this task with patient-level splitting to prevent information leakage. To better evaluate this task beyond surface-level text similarity, we further develop change-aware metrics specifically designed to capture clinically meaningful longitudinal changes, and conduct an independent physician validation to assess the reliability of the synthesized references and event extraction pipeline. We also compare direct paired-CT reasoning with an indirect two-stage pipeline that first generates single-timepoint reports and then performs textual differencing. Finally, we propose DeltaMed, a baseline model for direct paired-CT difference reporting, and train it on the benchmark training set. Together, these contributions lay the groundwork for temporally aware medical foundation models that better reflect real-world longitudinal clinical reasoning.

cs.CL↗
arXiv2026

Nanohertz Gravitational-Wave Constraints on Supermassive Binary Black Holes at Cosmic Dawn

Standard continuous gravitational-wave searches with pulsar timing arrays (PTAs) neglect cosmological redshift, restricting their applicability to the local Universe. We introduce a redshift-aware PTA framework and apply it to the Parkes PTA Data Release 3, deriving the first direct constraints on supermassive binary black holes (SMBBHs) at the cosmic dawn. Evaluating our limits across a broad redshift range, we observationally establish the non-monotonic mass-redshift exclusion boundary driven by the theoretical ``redshift bias", demonstrating how PTAs can effectively probe sources at extreme distances. Redshift-aware targeted searches toward high-redshift systems, including the ultraluminous quasar J0100+2802 ($z=6.327$) and the JWST-discovered galaxy JADES-GS-z14-0 ($z=14.32$), rigorously exclude SMBBHs with chirp masses $\mathcal{M}_c \gtrsim 10^{10} M_\odot$ across the nanohertz frequency band. Finally, we demonstrate that high-redshift binaries can be robustly detected and localized, allowing for the accurate measurement of their intrinsic properties. Our results provide the tightest constraints to date on SMBBHs at $z > 6$ and establish a practical framework for probing early-Universe black hole assembly.

astro-ph.HE↗