Search arXiv⌕ Search

arXiv subjects

Hao Jiang

Publications and source records attributed to Hao Jiang.

At least 73 records · Page 4Linked to original sources
arXiv2026

OmniaBench: Benchmarking General AI Agents Across Diverse Scenarios

Large language models are increasingly evolving from text generators into general agents capable of understanding user requests, invoking external tools, and completing complex tasks through interaction. However, existing agent benchmarks often focus on limited scenarios, tool ecosystems, or interaction formats, making it difficult to systematically characterize model capabilities across heterogeneous application settings. We introduce OmniaBench, a benchmark for evaluating general agents across diverse scenarios with explicit state spaces. We derive application-oriented scenario knowledge from app stores, product documents, industry resources, Web retrieval, and human refinement, forming a hierarchical taxonomy that spans ToC, ToB and ToE with 90 level-1 and 354 level-2 domains. Based on this taxonomy, we construct executable environments and synthesize single-turn and multi-turn tasks through four complementary routes: DAG, DAG-S, Solver, and Program. OmniaBench further introduces a ten-dimensional capability taxonomy and eight compositional atomic difficulty factors to support fine-grained evaluation and analysis. The resulting dataset contains 1,431 tasks, together with a challenging subset of 644 tasks designed to reduce evaluation cost and mitigate potential contamination of the full set after public release. The bench presents substantial challenges to current frontier models, with even Claude-Sonnet-5 and GPT-5.6-Sol achieving Overall Pass@1 scores of only 58.54 and 57.14, respectively. Further analyses reveal clear differences across domains and capabilities, as well as persistent limitations in planning, constraint maintenance, and adaptive correction. OmniaBench provides a broad and diagnostic benchmark for characterizing the capability boundaries of general agents.

cs.CL↗
arXiv2026

HANDFUL: Sequential Grasp-Conditioned Dexterous Manipulation with Resource Awareness

Dexterous robot hands offer rich opportunities for multifunctional manipulation, where a robot must execute multiple skills in sequence while maintaining control over previously grasped objects. Most prior work in dexterous manipulation focuses on single-object, single-skill tasks. In contrast, our insight is that many sequential tasks require resource-aware grasps that conserve fingers for future actions. In this paper, we study sequential grasp-conditioned dexterous manipulation, where a robot first grasps an object and then performs a second, distinct manipulation subtask while preserving the initial grasp. We introduce HANDFUL, a learning framework that models finger usage as a limited resource and encourages exploration of resource-aware grasps through finger-level contact rewards. These grasps are subsequently selected for downstream tasks via curriculum-based policy learning. We further propose HANDFUL-Bench, a simulation benchmark that introduces sequential dexterous manipulation tasks across multiple secondsubtask objectives, including pushing, pulling, and pressing, under a shared grasp-conditioned setup. Extensive simulation results demonstrate that prioritizing resource-aware grasps improves second-subtask success and robustness compared to a baseline that greedily optimizes the initial grasp before attempting the second subtask. We additionally validate our approach on a real dexterous LEAP hand. Together, this work establishes resource-aware grasp planning as a key principle for multifunctional dexterous manipulation. Supplementary material is available on our website: https://handful-dex.github.io.

cs.RO↗
arXiv2026

Alleviating Sparse Rewards by Modeling Step-Wise and Long-Term Sampling Effects in Flow-Based GRPO

Deploying GRPO on Flow Matching models has proven effective for text-to-image generation. However, existing paradigms typically propagate an outcome-based reward to all preceding denoising steps without distinguishing the local effect of each step. Moreover, current group-wise ranking mainly compares trajectories at matched timesteps and ignores within-trajectory dependencies, where certain early denoising actions can affect later states via delayed, implicit interactions. We propose TurningPoint-GRPO (TP-GRPO), a GRPO framework that alleviates step-wise reward sparsity and explicitly models long-term effects within the denoising trajectory. TP-GRPO makes two key innovations: (i) it replaces outcome-based rewards with step-level incremental rewards, providing a dense, step-aware learning signal that better isolates each denoising action's "pure" effect, and (ii) it identifies turning points-steps that flip the local reward trend and make subsequent reward evolution consistent with the overall trajectory trend-and assigns these actions an aggregated long-term reward to capture their delayed impact. Turning points are detected solely via sign changes in incremental rewards, making TP-GRPO efficient and hyperparameter-free. Extensive experiments also demonstrate that TP-GRPO exploits reward signals more effectively and consistently improves generation. Demo code is available at https://github.com/YunzeTong/TurningPoint-GRPO.

cs.CV↗
arXiv2026

Finite-Blocklength ISAC Multiple Access: A Source-Channel Coding Perspective

Future networks must serve massive populations of devices that sense and communicate simultaneously under short-packet constraints, yet the fundamental limits of integrated sensing and communication (ISAC) in the finite-blocklength multiple-access regime remain largely undiscovered. This paper closes this gap from a source-channel coding perspective. We prove that satisfying a sensing-distortion constraint is information-theoretically equivalent to a source-coding requirement, which collapses sensing and communication into the joint recovery of a single effective payload within a coded multiple-access framework. Building on this equivalence, we derive a finite-blocklength achievability bound together with a Fano-sum many-user converse and a genie-aided single-user converse, yielding a tight characterization of the minimum energy per bit and the rate-sensing tradeoff. Numerical results reveal that the energy price of sensing fidelity grows almost linearly in dB per decade of distortion tightening and is significantly amplified by the multiple-access load, and that joint encoding of the effective payload strictly outperforms an optimized orthogonal two-phase scheme, demonstrating a genuine integration gain of ISAC at finite blocklength.

cs.IT↗
arXiv2026

Occluding the Solution Space: Planner-Agnostic Adversarial Attacks on Tolerance-Aware Manipulation

Adversarial attacks on motion planning are crucial for evaluating and quantifying the intrinsic robustness of robotic manipulation. However, existing approaches are typically limited by restrictive exact-pose objectives and their reliance on planner-in-the-loop queries. To address these limitations, we propose a planner-agnostic attack framework for tolerance-aware manipulation. Our approach shifts the evaluation paradigm to task-level feasibility over goal regions, efficiently inserting adversarial obstacles without requiring oracle access to the victim system. Offline, we characterize the robot's intrinsic workspace capabilities via a kinematic occupancy heatmap, which encodes the density of feasible trajectories and robustness priors without invoking a specific planner. Online, we formulate the attack as a budgeted maximum-coverage optimization, strategically deploying obstacles subject to explicit geometric constraints to occlude the solution space. Extensive experiments across simulation and real-world scenarios demonstrate that our method reliably induces planning failures, significantly outperforming planner-in-the-loop baselines in both computational efficiency and attack efficacy.

cs.RO↗
arXiv2026

B2X Networks: Joint Design of Communication and Control for Embodied Intelligence

This article proposes the concept of \emph{brain-body-to-everything (B2X)} networks to facilitate the integration of wireless networks and embodied intelligence. In this framework, the \emph{brain} refers to the intelligence functions for reasoning, planning, and decision-making, the \emph{body} denotes the physical embodied agent that senses and acts in the real world, and \emph{X} represents the surrounding ecosystem involved in the brain-body interaction loop. Two B2X architectures with \emph{distributed} and \emph{centralized} brains are introduced to characterize different placements of intelligence across the body, base station, and core network. The uplink and downlink designs of B2X networks are then discussed under a representative base-station-side brain setting. For the uplink, communication is redesigned for B2X state acquisition under event urgency, sensing volume, and simultaneous multi-body access. For the downlink, communication is redesigned to coordinate command delivery and conventional service under shared radio resources. Based on these uplink and downlink considerations, a communication-control Pareto boundary is further used to characterize the loop-level trade-off between wireless transmission performance and control quality in B2X networks. Finally, several open research problems are discussed to guide future B2X network design.

eess.SP↗
arXiv2026

Fluid-Antenna-Aided Active User Detection With 1D-CNN Channel Reconstruction for Unsourced Random Access

In this paper, we investigate the application of fluid antenna systems (FAS) for active user detection (AUD) in unsourced random access (URA). A channel reconstruction method based on a one-dimensional convolutional neural network (1D-CNN) is proposed to effectively learn the nonlinear mapping from partial channel observations to the full channel vector. Furthermore, the reconstructed channel information is exploited to improve AUD performance via port selection. Simulation results demonstrate that the proposed 1D-CNN channel reconstructor significantly outperforms traditional methods under varying pilot lengths, achieving superior normalized mean squared error (NMSE) performance. Additionally, the reconstructed channel substantially reduces the AUD error rate compared with conventional approaches relying on traditional antenna configurations.

cs.IT↗
arXiv2026

Peak Sidelobe Suppression in Planar Fluid Antenna Array

Fluid antenna systems (FAS) have emerged as a promising technology for next-generation wireless communications, offering inherent reconfigurability and spatial adaptability. A distinctive and practically consequential property of fluid antenna arrays (FAAs) is their geometric diversity: by dynamically activating different subsets of spatially distributed ports across a dense discrete grid, a FAA can reconfigure its effective aperture geometry on demand, thereby unlocking unprecedented spatial degrees of freedom for radiation pattern synthesis. Exploiting such geometric flexibility, this paper investigates peak sidelobe level (PSLL) minimization in sparse planar FAAs through enhanced heuristic optimization. Specifically, an improved genetic algorithm (IGA) is proposed to determine the optimal port activation pattern that minimizes the PSLL under strict sparsity constraints. The proposed IGA incorporates tournament selection, adaptive operator probabilities, a hybrid crossover scheme, multi-point mutation, and an elite-pool preservation strategy to improve both convergence speed and solution quality. Simulation results demonstrate that the IGA significantly outperforms the canonical GA (CGA) in convergence behavior and final PSLL performance, achieving a 4.45 dB reduction in sidelobe levels while maintaining a comparable mainlobe width.

cs.IT↗
arXiv2026

Pinching Antennas-Assisted Sensing: A Ziv-Zakai Bound (ZZB) Perspective

The sensing capability of the pinching-antenna system (PASS) is analyzed from a Ziv-Zakai bound (ZZB) perspective, motivated by the sensing ambiguity arising from the multimodal observation model inherent to PASS. In comparison to other Bayesian sensing bounds, the ZZB provides a lower bound on the mean-squared error (MSE) across a broad range of signal-to-noise ratios (SNRs) and accounts for ambiguity in the likelihood functions. First, an observation model is developed for an uplink sensing scenario where a single sensing target transmits uplink pilots to a single-waveguide PASS receiver equipped with multiple pinching antennas (PAs). Building on this model, general ZZB expressions are derived for arbitrary prior distributions of the target's position, and are then specialized to the Gaussian and uniform cases. Second, the asymptotic ZZBs in low- and high-SNR regimes are characterized, and the relationship between the ZZBs and the conventional Bayesian Cramér-Rao bound (BCRB) is further studied by introducing the concept of an ambiguity function. Furthermore, to reduce the high computational complexity of direct evaluation of the ZZB, SNR-free and SNR-aware surrogate objective functions are proposed to facilitate ZZB-based optimization for enhancing sensing performance. Numerical results demonstrate that: i) Compared with the BCRB, the ZZB provides a tight sensing performance lower bound over a wide range of SNRs, ii) the ambiguity-awareness of the ZZB can address the multimodality-induced ambiguity in sensing, thereby yielding a reliable lower bound on the MSE, and iii) the proposed surrogate objective functions enable effective ZZB minimization with a lower computational complexity.

eess.SP↗
arXiv2026

Fluid Antenna-assisted Unsourced ISAC Massive Access

Unsourced integrated sensing and communication (UNISAC) has emerged as a promising paradigm for supporting massive connectivity in 6G networks. However, existing approaches predominantly rely on fixed-position antennas at the base station (BS) and user equipment (UE). In uplink transmission with huge access density and limited resource budgets (i.e., finite blocklength, FBL), the fixed arrays are constrained by their physical aperture and static spatial sampling, which lead to severe multi-user interference and an unavoidable pilot collision error floor. To conquer the bottleneck derived from fixed-position physical constraint and utilize the abundant spatial diversity within compact space, this paper proposes a novel unsourced ISAC framework incorporating a fluid antenna system (FAS) at the user side. The proposed scheme exploits the positional flexibility of FAS to reconfigure the channel environment by continuously adjusting antenna ports in the spatial domain. Numerical results demonstrate that the proposed FAS-aided approach significantly reduces the per-user probability of error (PUPE) and enhances angle-of-arrival (AOA) sensing accuracy. Specifically, the proposed scheme provides a 40 dB capacity gain over traditional TDMA at 1000 active users. It should be noted that the FAS considered in this paper is only deployed at the transmitter. In our future work, we will try deploying FAS at both the transmitter and receiver.

cs.IT↗
arXiv2026

Democratizing and accelerating AI-driven pathology research through agentic intelligence

Computational pathology has advanced rapidly with the emergence of foundation models, yet widespread adoption remains limited by substantial technical complexity and programming requirements. Here we present PathLab, an autonomous agentic framework that translates natural-language research objectives into executable and validated computational pathology workflows through the structured composition of domain-specific skills and tools. By organizing workflow generation around reusable methodological modules, including data preprocessing, model development, evaluation and interpretation, PathLab enables studies to be specified at the level of scientific intent rather than implementation details. We evaluated PathLab across 12 public datasets spanning four representative task families: region-of-interest classification, whole-slide image classification, segmentation and survival prediction. Across all task categories, PathLab achieved non-inferior performance relative to expert implementations, while consistently enforcing semantic validity of user prompts and proactively rejecting incompatible workflow specifications prior to execution. In controlled user studies, PathLab substantially reduced the time required to generate executable analytical pipelines and enabled domain experts without programming experience to independently design, execute and evaluate computational pathology studies. Together, these results establish PathLab as a reliable interface between biomedical intent and computational execution, enabling computational pathology studies to be designed at the level of scientific questions rather than programming expertise. By lowering technical barriers to advanced AI methodologies, PathLab provides a foundation for the broader democratization of computational pathology.

cs.AI↗
arXiv2026

LLM Evolution as an Industry-Scale Ecosystem: A Lifecycle Perspective on Continual Learning

Continual learning capability is critical for Industrial LLMs, as deployed models must be continuously updated to meet evolving requirements and environments, rather than repeatedly retrained from scratch. However, most existing research focuses on improvements on static benchmarks, failing to capture real industrial needs. In this survey, we reformulate Industrial Continual Learning (ICL) for LLMs as a closed-loop update-and-release problem in a versioned ecosystem, where updates propagate hierarchically to industrial, application-specific models and LLM-powered applications, with capability inheritance and transfer across versions and model families. From this ecosystem perspective, we identify three core challenges: repeated adaptation erodes model plasticity, foundation-model upgrades break capability inheritance, and long-term sustainability is constrained by deployment requirements. We then organize the technical landscape of ICL around five lifecycle design principles: preserving plasticity headroom, treating upgrades as capability transfer, enabling trustworthy continual reinforcement learning, making training recipes self-optimizing, and building accountability as a base layer for long-term iteration. For each principle, we synthesize representative technical directions. Finally, we evaluate the maturity of each principle and its technical components via an evidence-based lens, identify key gaps hindering real-world deployment, and outline a practical ICL deployment blueprint and a pathway for feeding industrial realities back into academic research.

cs.LG↗
arXiv2026

PRISMR: Overcoming Parse Collapse in Multimodal Listwise Ranking via Parameterized Representation Internalization

Generative listwise ranking with Large Multimodal Models (LMMs) aims to capture global list context in a single forward pass, but its effectiveness degrades in long-context multimodal scenarios. We identify a recurring failure mode, parse collapse, where the autoregressive decoder produces fluent yet incomplete rankings by silently omitting candidates and terminating early. This failure stems from limited context utilization rather than simple formatting mistakes, making prompt engineering and constrained decoding insufficient. We propose PRISMR (Parameterized Representation Internalization for Semantic Multimodal Ranking), a framework that replaces transient in-context list processing with parametric structural conditioning. PRISMR uses a lightweight hypernetwork to encode multimodal candidates in parallel and generate item-specific LoRA weights, which are synthesized into an instance-specific adapter for a LMM. This paradigm enables more robust internalization of list structure while preserving the base model. We further introduce a large-scale multimodal review-ranking benchmark for evaluation. Experiments demonstrate that PRISMR substantially reduces parse collapse, improves listwise ranking performance, and transfers effectively across domains and instruction-tuned backbones.

cs.AI↗
arXiv2026

Beamforming Gain with Single-RF Movable Arrays

A single-radio-frequency (RF) movable array is investigated, in which all movable elements are driven by a single RF chain with equal amplitude and equal phase. The achievable beamforming gain enabled by antenna placement is analyzed. Linear beamforming gain scaling with the number of antennas is shown to be achievable in single-path channels, while coherent-combining conditions and aperture requirements are established for multipath channels. For multiuser transmission, the optimal max-min power allocation is derived in closed form, based on which an element-wise coordinate-search algorithm is developed for antenna placement design. Numerical results validate the analysis and reveal a fundamental tradeoff: beamforming gains can be achieved through antenna placement alone, but only at the expense of increased aperture resources.

eess.SP↗
arXiv2026

A Comprehensive Survey of Direct Preference Optimization: Datasets, Theories, Variants, and Applications

With the rapid advancement of large language models (LLMs), aligning policy models with human preferences has become increasingly critical. Direct Preference Optimization (DPO) has emerged as a promising approach for alignment, acting as an RL-free alternative to Reinforcement Learning from Human Feedback (RLHF). Despite DPO's various advancements and inherent limitations, an in-depth review of these aspects is currently lacking in the literature. In this work, we present a comprehensive review of the challenges and opportunities in DPO, covering theoretical analyses, variants, relevant preference datasets, and applications. Specifically, we categorize recent studies on DPO based on key research questions to provide a thorough understanding of DPO's current landscape. Additionally, we propose several future research directions to offer insights on model alignment for the research community. An updated collection of relevant papers can be found on https://github.com/Mr-Loevan/DPO-Survey.

cs.AI↗
arXiv2026

LoomVideo: Unifying Multimodal Inputs into Video Generation and Editing

Developing unified video generation and editing models capable of interpreting interleaved multimodal inputs is a promising yet challenging frontier field. Existing unified frameworks predominantly rely on massive models (typically 13B parameters or more) and incorporate source video conditions for editing by concatenating sequence tokens. This concatenation inevitably doubles the sequence length, quadrupling the computational complexity of the self-attention mechanism and introducing prohibitive overhead. To address these bottlenecks, we present LoomVideo, a highly efficient 5B-parameter unified architecture for both video generation and editing. LoomVideo replaces the standard text encoder with a Multimodal Large Language Model (MLLM) and employs Deepstack injection mechanism to align multi-layer MLLM features with the Diffusion Transformer (DiT). Crucially, we introduce a zero-overhead Scale-and-Add conditioning approach for video editing. By scaling and directly adding the clean source video latent to the noised target latent, this elegant design eliminates the need for token concatenation, drastically reducing computational cost while maintaining robust capabilities for complex, non-rigid edits. Furthermore, a Negative Temporal RoPE strategy is seamlessly integrated to handle multiple reference images. Extensive experiments demonstrate that our compact 5B model achieves state-of-the-art or highly competitive performance across comprehensive benchmarks, exhibiting exceptional superiority in e-commerce and fashion generation scenarios. Benefiting from the zero-overhead conditioning mechanism, LoomVideo achieves at least a 5.41x acceleration in inference speed compared to models of similar capabilities, paving the way for highly practical and efficient video foundation models.

cs.CV↗
arXiv2026

OneReason Technical Report

Generative recommendation models in the OneRec family have been widely deployed in many real-world services, such as short-video, live-streaming, advertising, and e-commerce. However, these generative models can only benefit from the scaling advantage, while their reasoning ability is hard to activate, since we cannot construct meaningful Chain-of-Thought (CoT) sequences consisting of itemic tokens only. Inspired by the success of the reasoning-style ``think before answer'' paradigm in the LLM field, we conduct preliminary studies (i.e., OneRec-Think, OpenOneRec) to explore reasoning capability in generative recommendation. Nevertheless, we notice an unexpected phenomenon: the thinking mode does not show advantages over the non-thinking mode. Drawing insights from recent findings on CoT robustness in multi-modal language models, we argue that effective reasoning in recommendation rests on two factors: perception, the ability to ground itemic tokens in their underlying language semantics, and cognition, the ability to reorganize a user's behavior sequence into coherent latent interest points. We therefore propose OneReason, which includes: (1) strong itemic token perception in pre-training, (2) a three-level cognition-enhanced CoT format for recommendation tasks in SFT, and (3) a specialize-then-unify training recipe in RL to enhance the thinking ability.

cs.IR↗
arXiv2026

Three-dimensional density and air-rock interface reconstruction with muography: Application to the TianQin tunnel

Muography is a non-invasive imaging technique that uses cosmic-ray muons, commonly divided into transmission (absorption) and scattering muography. For transmission muography, the inversion algorithm critically determines reconstruction quality. However, widely used schemes may produce smearing artifacts when measurement locations are limited and data are sparse. We develop an optimized Metropolis--Hastings (M--H) algorithm that mitigates smearing and retrieves sharper, more accurate density distributions without auxiliary data. Additionally, we implement an inverse distance weighting (IDW) approach to reconstruct the air--rock interface from muon measurements. The optimized M--H algorithm is applied in Monte Carlo simulations and applied to field data from the TianQin Tunnel experiment using the MuGrid-v2 detector. The IDW-reconstructed air--rock interface is validated against Light Detection and Ranging (LiDAR) measurements. In simulations, the optimized M--H algorithm improves high-density anomaly detection precision from $42\%$ to $100\%$ at threshold $5.1\,\mathrm{g/cm^3}$, with gains of $6\%$ to $42\%$ across other threshold and low-density scenarios, together with the TianQin Tunnel reconstructions, these results demonstrate the effectiveness of the proposed approach.

physics.ins-det↗