arXiv subjects

Zhe Hu

Publications and source records attributed to Zhe Hu.

At least 19 recordsLinked to original sources

arXiv2026

DREAM Technical Report

Industrial recommender systems commonly use cascaded retrieval, ranking, and re-ranking pipelines. Although efficient, these pipelines fragment information and objectives across modules, rely on rigid rules, and have limited awareness of real-time intent, leaving session-level shifts among browsing, comparison, and purchase insufficiently addressed. We present DREAM (Developing Recommender Engine with Agentic Methods), an autonomous optimization control architecture that adds a perception-aware, orchestrable, and auditable policy layer atop existing pipelines without replacing them. DREAM has two core components. First, a three-tier Intent Engine fuses on-device signals into structured L0/L1/L2 intent representations; its edge-cloud trigger chain reduces reporting volume to approximately 8.7%. Second, a Meta Engine uses a MetaModel for layered M1-to-M2-to-M3 reasoning: intent summarization, strategy planning informed by Strategy Memory, and parameter translation. It dispatches the resulting parameters through a unified outlet with safety guardrails. A Reward Dual Loop continuously optimizes both components by combining offline simulation for strategy-space exploration with online feedback for outcome calibration, forming a cycle of generation, execution, evaluation, and experience accumulation. Large-scale A/B tests on Taobao's homepage feed show that re-ranking control alone improves IPV by 2.06%, Core IPV by 2.39%, and GMV by 0.88%. Extending control to fine ranking raises these gains to 2.71%, 3.06%, and 1.31%, respectively, while consistently improving PV by more than 1%. These gains require neither replacement of pipeline models nor compromise of serving stability, supporting agentic meta-control as a viable paradigm for industrial recommendation.

cs.IR↗

arXiv2026

Capek 0.5: An Execution-Centric Vision-Language Model for Embodied Intelligence

Vision-language models are increasingly serving as the reasoning core of embodied agents. Robot execution is inherently iterative: each action reshapes the scene and physical state, continually renewing what must be perceived, reasoned about, and verified. Meeting these demands requires complementary capabilities that differ in supervision signals, prediction formats, and verification criteria. Existing approaches typically develop these capabilities against isolated, task-specific objectives, leaving open how they should be organized and integrated around execution as a whole. We present Capek 0.5, an embodied vision-language model built around an execution-centric capability taxonomy. Rather than organizing training by datasets or tasks, the taxonomy groups embodied capabilities according to their functional roles throughout execution and comprises four capability families: Spatial Reasoning, Temporal Understanding, Action Guidance, and State Verification. Each capability is first acquired by a dedicated specialist through reinforcement learning with verifiable rewards from a shared backbone, and the specialists are then consolidated into a single inference-time model through weight-space merging followed by routed policy-space distillation. We instantiate Capek 0.5 at the 2B and 35B-A3B scales and evaluate it from three complementary perspectives: comprehensive benchmark suites including Capek-StateBench, a new benchmark for state verification; a controlled study of capability retention from specialists to the unified model; and closed-loop evaluation in simulated embodied environments. Capek 0.5 improves the large majority of matched benchmark rows over its initialization, retains all four specialized capabilities in one checkpoint with quantified losses, and transfers to closed-loop embodied task execution.

cs.AI↗

arXiv2026

ARGUS: Theory-of-Mind Guided Argument Generation with Strategy-Aware Planning and Knowledge Grounding

Persuasive argument generation requires modeling audience beliefs, rhetorical strategies, and factual grounding. Despite recent advancements, existing methods remain largely audience-agnostic and fail to integrate strategy selection to improve persuasiveness. To bridge this gap, we propose Argus, an agent-based framework that operationalizes classical rhetoric for persuasive writing. At its core, a Theory-of-Mind (ToM) Reasoner constructs an explicit dual mental model of the audience's beliefs and values to guide downstream decisions. This representation conditions a component-aware planner that decomposes the argument into subtopics, assigns fine-grained rhetorical functions (logos, pathos, ethos), and triggers strategy-guided evidence retrieval at planning time. Finally, a refinement module iteratively targets and resolves multi-dimensional weaknesses without quality regression. We evaluate Argus across three diverse benchmarks using both automated pairwise Elo and LLM-as-judge metrics. Results show that Argus consistently outperforms strong baselines across multiple backbone models, achieving top rankings and the highest overall scores. Targeted simulation experiments further validate its effectiveness in shifting resistant audience stances.

cs.CL↗

arXiv2026

EvoHIL: Self-Evolving Reward and Flow-Matched Policy Optimization for Robust Human-in-the-Loop Reinforcement Learning

Human-in-the-loop reinforcement learning (HIL-RL) enables robots to learn contact-rich manipulation from limited real-world interaction, but deployment exposes three coupled limitations: static visual reward models fail under scene changes; independently sampled actions cause temporally inconsistent motion; and vision-based policies remain sensitive to appearance shifts. We present EvoHIL, a unified framework that adapts the reward model, action generator, and visual do main within a staged human-in-the-loop learning process. First, self-evolving reward (SER) adapts the success classifier from human-confirmed positives and provisional weak negatives. Second, Action Flow Stabilization (AFS) generates temporally coherent action chunks through flow matching, grounding policy updates in executed action prefixes and demonstrated behavior. Third, retention-aware offline fine-tuning replays relit interaction data while anchoring the AFS actor-critic to prior behavior, adapting the visual domain without additional robot interaction. Across six manipulation tasks on Franka FR3 and SO-101 arms under a controlled lighting shift, EvoHIL improves task success, agreement with human-confirmation labels, motion smoothness, and completion time relative to human-in-the-loop and imitation baselines.Project page: https://anonymous4366.github.io/EvoHIL/

cs.RO↗

arXiv2026

Computational Humor with Multimodal LLMs: Methods, Datasets, Evaluation, and Challenges

Multimodal humor in memes, cartoons, and comics remains difficult for AI systems because intended meaning depends on non-literal mechanisms, shared cultural knowledge, and communicative intent rather than literal scene description. This survey focuses on visual humor understanding in single-image and multi-panel artifacts, while treating humor generation as an emerging downstream frontier. We position the literature against prior humor, sarcasm, and general MLLM surveys and organize it using a capability-centric hierarchy spanning recognition, interpretation and reasoning, and generation. Under this lens, we synthesize benchmark design, evaluation protocols, and modeling paradigms, tracing the field's shift from task-specific fusion models to large-model approaches based on multimodal alignment, evidence-grounded reasoning, and controlled generation. We conclude by highlighting the main barriers to progress: shortcut-prone evaluation, limited cultural and narrative coverage, weak evidence grounding, and unresolved safety and ownership concerns.

cs.CL↗

arXiv2026

Clinical Harness for Governable Medical AI Skill Ecosystems

Medical AI remains organized around isolated models, whereas care requires accountable capabilities that persist across time. We define clinical AI skills and propose the Clinical Harness, a runtime governance architecture that registers, orchestrates, constrains and monitors them. Using osteoporosis as an exemplar, we show how knowledge-driven, data-driven and physics-enhanced skills can support lifecycle care and provide a governed substrate for future medical agents.

cs.AI↗

arXiv2026

VIABLE: A Visually Impaired Assistance Benchmark for VLM-as-a-Judge Evaluation

AI-based Visually Impaired Assistance (VIA) remains challenging, largely due to the high cost of human evaluation. The VLM-as-a-Judge paradigm may offer a promising alternative, although it has mostly been studied in general domains. We therefore ask whether such judges can be trusted for VIA tasks. To investigate this question, we introduce VIABLE (Visually Impaired Assistance Benchmark for VLM-as-a-Judge Evaluation), the first benchmark for VLM-as-a-Judge evaluation in VIA. VIABLE contains over 300K judgment samples across three scenarios and introduces an Effectiveness--Impartiality--Stability framework with a 12-mode failure taxonomy. Based on VIABLE, our systematic study of seven judges across different model scales shows that existing models are largely unreliable across all evaluation axes. The strongest judge, GPT-5.4, achieves only 52.6% single-failure diagnostic accuracy, yet exhibits the highest self-preference rate at 94.2%; while open-source judges are strongly biased and adversarially fragile. To address these issues, we propose VIA-Judge-Agent, a model-agnostic inference-time harness that augments judges with visual evidence extraction and a taxonomy-guided workflow. It enables positive improvements in diagnostic accuracy and downstream VIA responses more preferred by BLV users. Data and code are available at: https://github.com/YiyiyiZhao/VIABLE

cs.CL↗

arXiv2026

Mass distribution of neutron stars in binary systems

It is known that the mass distribution of the known neutron stars (NSs) exhibits a bimodal pattern. The origin of this distribution remains a subject of debate. We constructed a super-Eddington accretion model for accreting neutron stars and investigated the mass growth and distribution of these stars using the population synthesis method. We find, in our model, the mass growth of NSs depends on the binary orbital period and the mass of the donor star. Our results can successfully account for the bimodal distribution of NS masses. The peak distribution of NS masses at around ~ 1.8 Msun primarily originates from NS binary systems where the donor star mass is less than ~ 1.6 Msun and the orbital period is shorter than 20 days; while, NS systems that may undergo common envelope evolution and these NSs can account for the mass peak at 1.4 Msun.

astro-ph.SR↗

arXiv2026

RoHIL: Robust Human-in-the-Loop Robotic Reinforcement Learning Against Illumination Variations

Human-in-the-loop reinforcement learning systems achieve near-perfect success on the workstation where they are trained, but collapse when the same robot is moved to a workstation a few meters away due to shifts in the visual input distribution caused by new lamp positions and window light. Re-collecting demonstrations and re-running HIL on every workstation is incompatible with deployment, and naively fine-tuning on shifted-light data triggers catastrophic forgetting of the source workstation. To close this cross-domain gap, we present RoHIL, an offline fine-tuning framework that uses no extra real-robot interaction. RoHIL combines (i) a world-model-based image relighter that re-synthesises the visual stream of source-workstation trajectories under multiple virtual HDRI environments, leaving actions and rewards real; (ii) Illumination-Retention Replay (IRR), a data-level anti-forgetting mechanism that interleaves relit adaptation transitions with original-light retention transitions to preserve source-workstation Bellman coverage; and (iii) an anchored Bellman-actor regulariser that constrains representation and policy drift from the original source-workstation policy. Across four real-robot manipulation tasks under significant cross-workstation illumination variations, RoHIL substantially improves shifted-light performance where standard HIL-RL collapses, while preserving source-workstation performance, eliminating the need to re-collect data and retrain for every new workstation and environment. Project page: https://anonymous4365.github.io/RoHIL/

cs.RO↗

arXiv2026

An evaluation framework for sparse 4D (3D + time) imaging reconstruction via bootstrapped cross-validation

Four-dimensional (4D; 3D + time) microscopic imaging has emerged as a powerful technique for investigating dynamic phenomena in complex systems, enabling direct visualization of structural evolution in space and time. However, when pushing the limits of spatiotemporal resolution, most time-resolved imaging techniques yield inherently sparse 4D datasets. While deep learning-based reconstruction methods have shown promise in reconstructing 4D from sparse spatiotemporal measurements, a practical approach for evaluating their performance in the absence of a 4D reference has, to the best of our knowledge, been lacking. Here, we present a bootstrapped cross-validation framework that estimates reconstruction performance by quantifying correlations between reconstructions generated from independently sampled subsets of the acquired data, as inspired by the 3D validation strategy in cryo-electron microscopy, where reconstructions from split datasets are compared to assess resolutions. This enables both qualitative and quantitative assessment in the absence of ground truth. We investigate two representative scenarios with sparse and ultra-sparse X-ray datasets and validate this approach using 4D-ONIX, a 4D deep-learning reconstruction method, on simulated water droplet collision experiments. The proposed approach provides a reference-free framework for performance estimation and support for better-informed experimental strategies across a wide range of ultrafast imaging applications.

eess.IV↗

arXiv2026

Overcoming Dynamics-Blindness: Training-Free Pace-and-Path Correction for VLA Models

Vision-Language-Action (VLA) models achieve remarkable flexibility and generalization beyond classical control paradigms. However, most prevailing VLAs are trained under a single-frame observation paradigm, which leaves them structurally blind to temporal dynamics. Consequently, these models degrade severely in non-stationary scenarios, even when trained or finetuned on dynamic datasets. Existing approaches either require expensive retraining or suffer from latency bottlenecks and poor temporal consistency across action chunks. We propose Pace-and-Path Correction, a training-free, closed-form inference-time operator that wraps any chunked-action VLA. From a single quadratic cost, joint minimization yields a unified solution that decomposes orthogonally into two distinct channels. The pace channel compresses execution along the planned direction, while the path channel applies an orthogonal spatial offset, jointly absorbing the perceived dynamics within the chunk window. We evaluate our approach on a comprehensive diagnostic benchmark MoveBench designed to isolate motion as the sole controlled variable. Empirical results demonstrate that our framework consistently outperforms state-of-the-art training-free wrappers and dynamic-adaptive methods and improves success rates by up to 28.8% and 25.9% in absolute terms over foundational VLA models in dynamic-only and static-dynamic mixed environments, respectively.

cs.RO↗

arXiv2026

Goal2Skill: Long-Horizon Manipulation with Adaptive Planning and Reflection

Recent vision-language-action (VLA) systems have demonstrated strong capabilities in embodied manipulation. However, most existing VLA policies rely on limited observation windows and end-to-end action prediction, which makes them brittle in long-horizon, memory-dependent tasks with partial observability, occlusions, and multi-stage dependencies. Such tasks require not only precise visuomotor control, but also persistent memory, adaptive task decomposition, and explicit recovery from execution failures. To address these limitations, we propose a dual-system framework for long-horizon embodied manipulation. Our framework explicitly separates high-level semantic reasoning from low-level motor execution. A high-level planner, implemented as a VLM-based agentic module, maintains structured task memory and performs goal decomposition, outcome verification, and error-driven correction. A low-level executor, instantiated as a VLA-based visuomotor controller, carries out each sub-task through diffusion-based action generation conditioned on geometry-preserving filtered observations. Together, the two systems form a closed loop between planning and execution, enabling memory-aware reasoning, adaptive replanning, and robust online recovery. Experiments on representative RMBench tasks show that the proposed framework substantially outperforms representative baselines, achieving a 32.4% average success rate compared with 9.8% for the strongest baseline. Ablation studies further confirm the importance of structured memory and closed-loop recovery for long-horizon manipulation.

cs.RO↗

arXiv2026

4D reconstruction of alumina laser melt pools at 25 kHz via operando X-ray multi-projection imaging

Advancing additive manufacturing, e.g., laser powder-bed fusion (LPBF), requires resolving rapid processes such as melt-pool dynamics and keyhole evolution in 4D (3D + time). Operando X-ray tomography is a state-of-the-art approach for 4D characterization, but its temporal resolution is fundamentally constrained by the sample rotation speed, limiting achievable 4D imaging rates and preventing the resolution of these fast phenomena. Here we present rotation-enabled X-ray Multi-Projection Imaging (rotation-XMPI), which captures three angularly resolved projections per time step and thereby decouples temporal resolution from the sample rotation speed. Combined with a self-supervised deep-learning reconstruction framework for multi-angle inputs, rotation-XMPI enables high-fidelity 4D imaging at unprecedented speed. We demonstrate the approach in an operando alumina laser-remelting experiment at MAX IV using three beamlets combined with 25 Hz sample rotation. Rotation-XMPI resolves melt-pool morphology and keyhole evolution; in contrast, conventional and limited-angle tomography remain rotation-limited, and motion blur prevents resolving these dynamics. Overall, rotation-XMPI delivers a 250-fold increase relative to state-of-the-art melt-pool imaging, effectively achieving 25,000 reconstructed volumes per second. This method establishes a practical route to scalable ultrafast 4D imaging for additive manufacturing and other materials processes.

physics.optics↗

arXiv2026

4D Synchrotron X-Ray Multi Projection Imaging (XMPI) for studying multiphase flow dynamics and flow instabilities in porous networks

Flow instabilities such as Haines jumps in porous media are common phenomena that occur on sub-second timescales. X-rays are particularly suitable for investigating these processes because they provide non-destructive three-dimensional insight into the network structure and the liquid distribution within porous media. Studying imbibition events in four dimensions (three spatial dimensions plus time) is inherently challenging with conventional tomography because the required rapid sample rotation imposes significant centrifugal forces that alter the flow. Here, we demonstrate synchrotron X-ray multi-projection imaging (XMPI) to capture four-dimensional flow in an additively manufactured, homogeneous spherical pore network at 1.3 $\mu$m effective pixel size and 50 Hz temporal resolution without the need for high rotational speeds. This enables in situ visualization of non-repeatable pore-scale events in both space and time, a capability unachievable with classical X-ray tomographic approaches. We compare the results to Shan-Chen multiphase Lattice Boltzmann simulations performed on the same geometry, finding both qualitative agreements and systematic differences in filling sequences and timescales. These discrepancies expose key limitations of current simulation methods in representing contact-line dynamics and realistic boundary conditions limitations that XMPI can directly overcome. By enabling high-resolution, real-time imaging of flow instabilities in opaque porous media, synchrotron XMPI provides a unique platform that bridges the gap between pore-scale experiments and simulations.

physics.flu-dyn↗

arXiv2026

HSC-VLA: Hierarchical Scene-Clearing for Robust Bimanual Manipulation in Dense Clutter

Modern Vision--Language--Action models often suffer from critical instruction-following failures in high-density manipulation environments, where task-irrelevant visual clutter dilutes attention, corrupts grounding, and substantially degrades performance in complex long-horizon scenarios. To overcome the representation bottleneck of monolithic end-to-end architectures, we propose HSC-VLA, a hierarchical framework that decouples high-level visual-semantic reasoning from low-level, high-frequency sensorimotor execution through an explicit scene-clearing abstraction. HSC-VLA employs a high-level Brain to decompose long-horizon tasks and to generate task-specific scene masks that preserve task-relevant geometry while suppressing distractors. The filtered observations are then passed to a low-level Cerebellum, a diffusion-based policy that performs bimanual manipulation using only mask-filtered vision and proprioception. Extensive experiments in densely cluttered supermarket shelves demonstrate that HSC-VLA achieves 86.7\% aggregate success under high-density clutter, surpassing the best monolithic baseline ($\pi_0$-Full FT at 34.3\%) by 52.4\%. HSC-VLA also exhibits strong long-horizon performance, reaching 72\% on clutter sorting and 66\% on restocking, demonstrating strong robustness and effective failure recovery in complex cluttered manipulation.

cs.RO↗

arXiv2026

Guided by the Plan: Enhancing Faithful Autoregressive Text-to-Audio Generation with Guided Decoding

Autoregressive (AR) models excel at generating temporally coherent audio by producing tokens sequentially, yet they often falter in faithfully following complex textual prompts, especially those describing complex sound events. We uncover a surprising capability in AR audio generators: their early prefix tokens implicitly encode global semantic attributes of the final output, such as event count and sound-object category, revealing a form of implicit planning. Building on this insight, we propose Plan-Critic, a lightweight auxiliary model trained with a Generalized Advantage Estimation (GAE)-inspired objective to predict final instruction-following quality from partial generations. At inference time, Plan-Critic enables guided exploration: it evaluates candidate prefixes early, prunes low-fidelity trajectories, and reallocates computation to high-potential planning seeds. Our Plan-Critic-guided sampling achieves up to a 10-point improvement in CLAP score over the AR baseline-establishing a new state of the art in AR text-to-audio generation-while maintaining computational parity with standard best-of-N decoding. This work bridges the gap between causal generation and global semantic alignment, demonstrating that even strictly autoregressive models can plan ahead.

cs.CL↗

arXiv2025

Constraint-Free Coherent Diffraction Imaging via Physics-Guided Neural Fields

CDI is a lensless imaging technique that enables atomic-resolution imaging of non-crystalline specimens and their dynamics. However, its broader implementation has been hindered by the instability and ill-posedness of its reconstruction process, known as phase retrieval, which relies heavily on handcrafted, object-specific constraints. To overcome the key limitations, we propose CDIP, a robust phase-retrieval framework that eliminates the need for such constraints by combining untrained coordinate-based neural fields for static and dynamic reconstructions and a physics-consistent forward model. We evaluate CDIP on simulated and experimental datasets that involve both static samples and dynamic processes, demonstrating that it substantially outperforms classical iterative algorithms and deep-learning baselines in terms of fidelity and stability. These results highlight a paradigm shift in both static and time-resolved CDI reconstruction, providing a broadly applicable framework for coherent imaging modalities such as ptychography and holography, across X-ray, electron, and optical probes.

physics.optics↗

arXiv2025

Exploring Scale Shift in Crowd Localization under the Context of Domain Generalization

Crowd localization plays a crucial role in visual scene understanding towards predicting each pedestrian location in a crowd, thus being applicable to various downstream tasks. However, existing approaches suffer from significant performance degradation due to discrepancies in head scale distributions (scale shift) between training and testing data, a challenge known as domain generalization (DG). This paper aims to comprehend the nature of scale shift within the context of domain generalization for crowd localization models. To this end, we address four critical questions: (i) How does scale shift influence crowd localization in a DG scenario? (ii) How can we quantify this influence? (iii) What causes this influence? (iv) How to mitigate the influence? Initially, we conduct a systematic examination of how crowd localization performance varies with different levels of scale shift. Then, we establish a benchmark, ScaleBench, and reproduce 20 advanced DG algorithms to quantify the influence. Through extensive experiments, we demonstrate the limitations of existing algorithms and underscore the importance and complexity of scale shift, a topic that remains insufficiently explored. To deepen our understanding, we provide a rigorous theoretical analysis on scale shift. Building on these insights, we further propose an effective algorithm called Causal Feature Decomposition and Anisotropic Processing (Catto) to mitigate the influence of scale shift in DG settings. Later, we also provide extensive analytical experiments, revealing four significant insights for future research. Our results emphasize the importance of this novel and applicable research direction, which we term Scale Shift Domain Generalization.

cs.CV↗