arXiv subjects

Zhen Wu

Publications and source records attributed to Zhen Wu.

At least 37 records · Page 2Linked to original sources

arXiv2026

OmniRetarget: Interaction-Preserving Data Generation for Humanoid Whole-Body Loco-Manipulation and Scene Interaction

A dominant paradigm for teaching humanoid robots complex skills is to retarget human motions as kinematic references to train reinforcement learning (RL) policies. However, existing retargeting pipelines often struggle with the significant embodiment gap between humans and robots, producing physically implausible artifacts like foot-skating and penetration. More importantly, common retargeting methods neglect the rich human-object and human-environment interactions essential for expressive locomotion and loco-manipulation. To address this, we introduce OmniRetarget, an interaction-preserving data generation engine based on an interaction mesh that explicitly models and preserves the crucial spatial and contact relationships between an agent, the terrain, and manipulated objects. By minimizing the Laplacian deformation between the human and robot meshes while enforcing kinematic constraints, OmniRetarget generates kinematically feasible trajectories. Moreover, preserving task-relevant interactions enables efficient data augmentation, from a single demonstration to different robot embodiments, terrains, and object configurations. We comprehensively evaluate OmniRetarget by retargeting motions from OMOMO, LAFAN1, and our in-house MoCap datasets, generating over 8-hour trajectories that achieve better kinematic constraint satisfaction and contact preservation than widely used baselines. Such high-quality data enables proprioceptive RL policies to successfully execute long-horizon (up to 30 seconds) parkour and loco-manipulation skills on a Unitree G1 humanoid, trained with only 5 reward terms and simple domain randomization shared by all tasks, without any learning curriculum.

cs.RO↗

arXiv2026

Recognize Your Orchestrator: An Entropy Dynamics Perspective for LLM Multi-Agent Systems

The transition from single-turn models to Multi-Agent Systems (MAS) promises enhanced problem-solving capabilities, yet the centralized orchestration topology remains a critical point of fragility. To analyze this, we propose a Mean-Field Entropy Dynamics framework, modeling the orchestration process as a system governed by the competing forces of task resolution and cumulative context loading. To facilitate validation, we introduce Inverse Workflow Generation (IWG), a multi-agent pipeline that synthesizes process-verifiable, high-complexity benchmarks with dense intermediate checkpoints. We demonstrate that our entropy dynamics model fits empirical trajectories, providing physically interpretable parameters that quantify system stability and performance collapse. Crucially, our analysis uncovers a ``Reasoning Trap": while reasoning-heavy models excel in isolated tasks, they frequently fail as orchestrators due to context squeezing. Elucidating the physical mechanisms underlying the Orchestrator and quantifying systemic uncertainty offers insights for the MASs' architectural design.

cs.AI↗

arXiv2026

Are LLMs Socially Adaptive? Contrasting Belief Evolution in Large Language Models and Humans

As large language models (LLMs) increasingly engage in complex social interactions, ensuring that their behaviors align with human ethical principles and intentions, known as value alignment, has become a critical scientific challenge. Existing benchmarks often rely on static assessments and fail to capture the longitudinal dynamics of decision-making or the latent cognitive processes driving agent behavior. In this work, we propose FairMindSim, a realistic simulation benchmark rooted in social psychology that evaluates alignment through continuous economic games. To move beyond black-box observations, we introduce the Belief-Reward Alignment Behavior Evolution Model (BREM), a probabilistic framework that formalizes decision-making as a dynamic trade-off between maximizing extrinsic rewards and upholding intrinsic beliefs. We conducted a large-scale comparative study involving 1,017 human participants and ten LLMs, including GPT-5 and Gemini-3-Pro. Our experimental results reveal a capability linked non linear empirical trend in the Third Party Punishment (TPP) game. Mid capability models exhibit rigid and algorithmic aggression that is characterized by over punishment, while frontier models show a convergence of restraint and a shift toward human like leniency as reasoning capabilities scale. Furthermore, using BREM, we decompose agents longitudinal decision dynamics and find that more advanced models better balance conflicting objectives by reducing belief action inconsistency. Our contributions provide a standardized protocol for psychological stress testing and an interpretable mechanism for analyzing the longitudinal evolution of AI alignment in controlled social dilemma settings.

cs.CE↗

arXiv2026

DetMesh-Gadep: Triangulated Surface Modeling and GPU-based Monte Carlo Efficiency Calibration of High-Purity Germanium Detectors

Sourceless efficiency calibration of high-purity germanium (HPGe) detectors can provide accurate detector-response information without experiments using radioactive calibration sources, offering advantages in both convenience and safety. In many practical implementations, this process is performed using Monte Carlo simulation; however, its performance is constrained by the accuracy of detector modeling, the operational complexity of simulation frameworks, and the computational-resource requirements associated with CPU-based parallelization. In this study, a complete detector modeling and simulation framework is proposed. The detector modeling program DetMesh can generate triangulated surface geometry from parameterized detector models, providing advantages in the representation of complex geometric boundaries. It incorporates standard geometric operations and a geometric library, and is lightweight with strong extensibility. The generated geometry is then input into Gadep, a GPU-based Monte Carlo computational kernel, to enable rapid simulation. For $1\times 10^8$ particles, a single RTX 4090 achieved a speedup factor of 13.53 compared with simultaneous computation using 60 CPU cores. The proposed framework has low implementation cost and broad applicability, providing a complete solution for refined detector modeling and calibration.

physics.ins-det↗

arXiv2026

Causal Evidence for Attention Head Imbalance in Modality Conflict Hallucination

Modality-conflict hallucination occurs when multimodal large language models (MLLMs) prioritize erroneous textual premises over contradictory visual evidence. To understand why visual evidence fails to prevail during generation, we take a mechanistic perspective and examine which internal components drive or resist this failure. We perform head-level causal analysis using path patching across five open-source MLLMs and identify two groups of attention heads with opposing causal roles: hallucination-driving heads and hallucination-resisting heads. We find a consistent asymmetry: driving effects are more broadly distributed and carry greater aggregate weight, whereas resisting effects concentrate in a small number of high-importance heads. Ablation experiments further confirm that these groups exert opposing effects during generation: distributed driving influence and localized resistance together form an imbalanced routing structure that biases generation toward the erroneous premise. Motivated by this finding, we propose MACI (Modality-conflict-Aware Causal Intervention), a conditional intervention that suppresses causally identified hallucination-driving heads only when conflict is detected. Across five MLLMs, MACI achieves the largest hallucination reduction among compared inference-time baselines on the MMMC benchmark with a favorable hallucination-accuracy trade-off, and transfers zero-shot to the SCI-SemanticConflict test.

cs.AI↗

arXiv2026

Deep Policy Iteration for High-Dimensional Mean-Field Games with Regenerative Reformulation

This paper develops a deep policy iteration method for high-dimensional finite-horizon mean-field games (MFG). We reformulate the game as a regenerative problem with deterministic cycles, which allows policy evaluation (PE), policy improvement (PI), and population measure estimation to be carried out cycle by cycle. Within this formulation, we approximate the population measure by a particle system and update it using a one-step random mapping induced by the Euler-Maruyama discretization of the state dynamics. This update transports a mini-batch of particles from one cycle to the next, avoiding sequential trajectory simulation over the entire time horizon at each iteration. The PE and PI subproblems are formulated through the relation between consecutive cycles, with adversarial training used for evaluation and averaged optimization used for improvement. The resulting method is efficient and scalable in high dimensions, as it avoids the direct solution of the coupled Hamilton-Jacobi-Bellman and Fokker-Planck system, the full simulation of trajectories to estimate the population measure, the explicit computation of conditional expectations in policy evaluation, and pointwise optimization in policy improvement. Numerical experiments demonstrate that the proposed method effectively handles dimensions up to 10,000.

math.NA↗

arXiv2026

Operational time-reversal symmetry for unital qubit channels

The Bayesian inverse of a quantum channel $\mathcal{E}$ is a channel $\mathcal{F}$ in the reverse direction of $\mathcal{E}$ that yields time-symmetric correlations for sequential measurements performed on open quantum systems. Such an operational form of time-reversal symmetry for open quantum systems is quite remarkable, as the dynamics of open quantum systems are inherently irreversible due to system-environment interactions. Similar to the Petz map, a Bayesian inverse $\mathcal{F}$ is defined with respect to a fiducial reference state $ρ$ for the channel $\mathcal{E}$. However, Bayesian inverses do not always exist, and it is often a non-trivial task to determine the set of states $ρ$ for which a Bayesian inverse of $\mathcal{E}$ exists. In this work, we solve the general problem of quantum Bayesian inversion for unital channels acting on a single qubit. Our analysis is streamlined by demonstrating that finding a Bayesian inverse for a unital qubit channel may be reduced to finding a Bayesian inverse of a Pauli channel, which is simply a mixture of unitary channels associated with the Pauli matrices. As such, we provide a complete description of when operational time-reversal symmetry is attainable for sequential measurements of a single qubit in the presence of unital noise.

quant-ph↗

arXiv2026

Perceptive Humanoid Parkour: Chaining Dynamic Human Skills via Motion Matching

While recent advances in humanoid locomotion have achieved stable walking on varied terrains, capturing the agility and adaptivity of highly dynamic human motions remains an open challenge. In particular, agile parkour in complex environments demands not only low-level robustness, but also human-like motion expressiveness, long-horizon skill composition, and perception-driven decision-making. In this paper, we present Perceptive Humanoid Parkour (PHP), a modular framework that enables humanoid robots to autonomously perform long-horizon, vision-based parkour across challenging obstacle courses. Our approach first leverages motion matching, formulated as nearest-neighbor search in a feature space, to compose retargeted atomic human skills into long-horizon kinematic trajectories. This framework enables the flexible composition and smooth transition of complex skill chains while preserving the elegance and fluidity of dynamic human motions. Next, we train motion-tracking reinforcement learning (RL) expert policies for these composed motions, and distill them into a single depth-based, multi-skill student policy, using a combination of DAgger and RL. Crucially, the combination of perception and skill composition enables autonomous, context-aware decision-making: using only onboard depth sensing and a discrete 2D velocity command, the robot selects and executes whether to step over, climb onto, vault or roll off obstacles of varying geometries and heights. We validate our framework with extensive real-world experiments on a Unitree G1 humanoid robot, demonstrating highly dynamic parkour skills such as climbing tall obstacles up to 1.25m (96% robot height), as well as long-horizon multi-obstacle traversal with closed-loop adaptation to real-time obstacle perturbations.

cs.RO↗

arXiv2026

CPI-C: Cool Planet Imaging Coronagraph on Chinese Space Station Survey Telescope

Cool Planet Imaging Coronagraph (CPI-C) on Chinese Space Station Survey Telescope (CSST) is proposed to direct image the cool planets around nearby solar-type stars (within 40 pc). The core scientific objective of CPI-C is to conduct high-contrast directly imaging surveys of exoplanets ranging in size from Neptune-like to Jupiter-like, located at separations of 0.5 to 5 AU from their host stars, and to perform systematic spectroscopic analysis of the detected planets through high-precision multi-band photometry. CPI-C employs a step-transmission apodization technique to suppress the diffraction noises from the telescope pupil and a precise phase correction technique to eliminate the speckle noises due to imperfections of the optical surfaces. The contrast requirement is better than $10^{-8}$ at an inner working angle (IWA) of $3-4λ/D$, in the visible wavelength from 600 nm to 900 nm. CPI-C will be the first space-based instrument capable of directly imaging the reflection light from the cool exoplanets in the visible wavelength enabling the measurement of key physical parameters such as the effective temperature, surface gravity, radius, mass, and other key parameters. The potential observation results will significantly contribute to further understand the formation and evolution mechanisms of planets, which will also lay a solid foundation for future confirmation of the Earth-twins in the next generation space flagship missions.

astro-ph.EP↗

arXiv2026

CogGen: A Cognitively Inspired Recursive Framework for Deep Research Report Generation

The autonomous synthesis of deep research reports represents a critical frontier for Large Language Models (LLMs), demanding sophisticated information orchestration and non-linear narrative logic. Current approaches rely on rigid predefined linear workflows, which cause error accumulation, preclude global restructuring from subsequent insights, and ultimately limit in-depth multimodal fusion and report quality. We propose CogGen, a Cognitively inspired recursive framework for deep research report Generation. Leveraging a Hierarchical Recursive Architecture to simulate cognitive writing, CogGen enables flexible planning and global restructuring. To extend this recursivity to multimodal content, we introduce Abstract Visual Representation (AVR): a concise intent-driven language that iteratively refines visual-text layouts without pixel-level regeneration overhead. We further present CLEF, a Cognitive Load Evaluation Framework, and curate a new benchmark from Our World in Data (OWID). Extensive experiments show CogGen achieves state-of-the-art results among open-source systems, generating reports comparable to professional analysts' outputs and surpassing Gemini Deep Research. Our code and dataset are available at https://github.com/NJUNLP/CogGen.

cs.MA↗

arXiv2026

The General Antiparticle Spectrometer (GAPS) Antarctic Balloon Payload

The General Antiparticle Spectrometer (GAPS) is an Antarctic stratospheric balloon mission designed to provide unmatched sensitivity to low-energy (<0.25 GeV/n) cosmic-ray antiprotons, antideuterons, and antihelium nuclei as signatures of dark matter. The distinctive GAPS particle identification technique relies on measuring the energy loss along the track of an incoming antinucleus as it slows down and is captured into an exotic atom, and then detecting the de-excitation X-rays and the nuclear annihilation products. This measurement is realized using a Tracker composed of more than 1000 custom silicon strip detectors and a plastic scintillator time-of-flight (TOF) system instrumenting more than 40m$^2$. Together, these subsystems provide the velocity and energy resolution, stopping power, particle tracking, and X-ray identification necessary to distinguish rare antinucleus signals from the abundant positive-nucleus backgrounds, all within the constraints of a high-altitude mission. A multi-loop capillary heat pipe system has been developed to maintain the tracker operating temperature with significant mass and power savings over a conventional pump-based system. The first GAPS science payload flew for 25 days during the 2025/26 NASA Antarctic balloon campaign. We detail the design, integration, and commissioning of the payload prior to flight.

astro-ph.IM↗

arXiv2026

SOAP: Enhancing Spatio-Temporal Relation and Motion Information Capturing for Few-Shot Action Recognition

High frame-rate (HFR) videos of action recognition improve fine-grained expression while reducing the spatio-temporal relation and motion information density. Thus, large amounts of video samples are continuously required for traditional data-driven training. However, samples are not always sufficient in real-world scenarios, promoting few-shot action recognition (FSAR) research. We observe that most recent FSAR works build spatio-temporal relation of video samples via temporal alignment after spatial feature extraction, cutting apart spatial and temporal features within samples. They also capture motion information via narrow perspectives between adjacent frames without considering density, leading to insufficient motion information capturing. Therefore, we propose a novel plug-and-play architecture for FSAR called Spatio-tempOral frAme tuPle enhancer (SOAP) in this paper. The model we designed with such architecture refers to SOAP-Net. Temporal connections between different feature channels and spatio-temporal relation of features are considered instead of simple feature extraction. Comprehensive motion information is also captured, using frame tuples with multiple frames containing more motion information than adjacent frames. Combining frame tuples of diverse frame counts further provides a broader perspective. SOAP-Net achieves new state-of-the-art performance across well-known benchmarks such as SthSthV2, Kinetics, UCF101, and HMDB51. Extensive empirical evaluations underscore the competitiveness, pluggability, generalization, and robustness of SOAP. The code is released at https://github.com/wenbohuang1002/SOAP.

cs.CV↗

arXiv2026

Coexisting topological hinges and 1D Rashba states in Bi$_{0.97}$Sb$_{0.03}$ revealed by the Josephson effect

Second-order topological insulating (SOTI) states in three-dimensional materials are helical one-dimensional hinge states. Inducing superconductivity in these states leads to gapless bound states, characterized by the 4$π$-periodic current-phase relation. Here, we provide evidence of the topologically protected hinge states in Dirac semimetal Bi$_{0.97}$Sb$_{0.03}$ nanoflakes by an unconventional interference pattern in a magnetic field, and the 4$π$-periodic supercurrent carried by these states via the suppressed first and third Shapiro steps. Tight-binding simulations confirm the presence of multiple hinge modes, supporting our interpretation of Bi$_{0.97}$Sb$_{0.03}$ as a prototypical designable SOTI platform. Quantum confinement effect is identified by a quasi-one-dimensional bulk transport, and the confined Rashba states are responsible for the broadened hinge states.

cond-mat.mes-hall↗

arXiv2026

$R^2$-CoD: Understanding Text-Graph Complementarity in Relational Reasoning via Knowledge Co-Distillation

Relational reasoning lies at the core of many NLP tasks, drawing on complementary signals from text and graphs. While prior research has investigated how to leverage this dual complementarity, a detailed and systematic understanding of text-graph interplay and its effect on hybrid models remains underexplored. We take an analysis-driven approach to investigate text-graph representation complementarity via a unified architecture that supports knowledge co-distillation (CoD). We explore five tasks involving relational reasoning that differ in how text and graph structures encode the information needed to solve that task. By tracking how these dual representations evolve during training, we uncover interpretable patterns of alignment and divergence, and provide insights into when and why their integration is beneficial.

cs.AI↗

arXiv2026

PRIMAL: Processing-In-Memory Based Low-Rank Adaptation for LLM Inference Accelerator

This paper presents PRIMAL, a processing-in-memory (PIM) based large language model (LLM) inference accelerator with low-rank adaptation (LoRA). PRIMAL integrates heterogeneous PIM processing elements (PEs), interconnected by 2D-mesh inter-PE computational network (IPCN). A novel SRAM reprogramming and power gating (SRPG) scheme enables pipelined LoRA updates and sub-linear power scaling by overlapping reconfiguration with computation and gating idle resources. PRIMAL employs optimized spatial mapping and dataflow orchestration to minimize communication overhead, and achieves $1.5\times$ throughput and $25\times$ energy efficiency over NVIDIA H100 with LoRA rank 8 (Q,V) on Llama-13B.

cs.AR↗

arXiv2026

The spatiotemporal Born rule is quasiprobabilistic

Contrary to general relativity, quantum theory treats space and time in fundamentally different ways. In particular, while joint probabilities associated with spacelike separated measurements are defined in terms of the Born rule, joint probabilities associated with measurements performed in sequence are defined in terms of the state-update rule. In this work, we show that one obtains a more unified perspective of space and time in quantum theory by embracing a quasiprobabilistic description of sequential measurements. More precisely, we show that there exists a unique \emph{pseudo}-density operator encoding canonical quasiprobabilities associated with sequential measurements in precisely the same manner that a density operator encodes joint probabilities associated with spacelike separated measurements, thus providing a natural extension of the Born rule into the temporal domain. As an application, we show how such a spatiotemporal Born rule combined in conjunction with a quantum Bayes' rule yields an operational notion of time-reversal symmetry for sequential measurements on an \emph{open} quantum system.

quant-ph↗

arXiv2026

Locomotion Beyond Feet

Most locomotion methods for humanoid robots focus on leg-based gaits, yet natural bipeds frequently rely on hands, knees, and elbows to establish additional contacts for stability and support in complex environments. This paper introduces Locomotion Beyond Feet, a comprehensive system for whole-body humanoid locomotion across extremely challenging terrains, including low-clearance spaces under chairs, knee-high walls, knee-high platforms, and steep ascending and descending stairs. Our approach addresses two key challenges: contact-rich motion planning and generalization across diverse terrains. To this end, we combine physics-grounded keyframe animation with reinforcement learning. Keyframes encode human knowledge of motor skills, are embodiment-specific, and can be readily validated in simulation or on hardware, while reinforcement learning transforms these references into robust, physically accurate motions. We further employ a hierarchical framework consisting of terrain-specific motion-tracking policies, failure recovery mechanisms, and a vision-based skill planner. Real-world experiments demonstrate that Locomotion Beyond Feet achieves robust whole-body locomotion and generalizes across obstacle sizes, obstacle instances, and terrain sequences.

cs.RO↗

arXiv2025

On small perturbations of coherent information

Quantum capacity quantifies the amount of quantum information that can be transmitted by a quantum channel with an arbitrary small probability of error. Mathematically, the quantum capacity is given by an asymptotic formula involving the one-shot quantum capacity of the associated channel, which, due to purely quantum effects such as superadditivity of one-shot quantum capacity, is rarely computable. The one-shot quantum capacity is mathematically characterized in terms of optimizing an entropic quantity referred to as coherent information over all possible input states of a channel, the computation of which also tends to be intractable due to the difficulty of optimizing the coherent information. In this work, we develop perturbative methods for analyzing the behavior of coherent information of a quantum channel with respect to small perturbations of the input state. By doing so, we are able to derive three general criteria for determining whether an input state yields suboptimal coherent information. We then show how our criteria yield sufficient conditions for superadditivity of one-shot quantum capacity, and also for detecting a positive gap between one-shot private capacity and one-shot quantum capacity. The utility of our criteria is illustrated through examples, which yield new results regarding the one-shot quantum capacity of qubit depolarizing channels, Pauli channels and dephrasure channels.

quant-ph↗