arXiv subjects

Jun Fang

Publications and source records attributed to Jun Fang.

At least 19 recordsLinked to original sources

arXiv2026

Reconfigurable Antenna-Enhanced ISAC: A Unified Perspective on Electromagnetic Adaptation

Integrated sensing and communication (ISAC) enables sensing and communication (S&C) functions to share time, frequency, and hardware resources, offering an efficient paradigm for next-generation wireless systems. However, conventional fixed-configuration antenna (FCA) architectures provide limited flexibility in adapting the physical electromagnetic (EM) response of the end-to-end wireless link. This article develops a unified EM-adaptation perspective on reconfigurable antenna (RA)-enhanced ISAC systems. We distinguish transceiver-side (TSc-side) and environment-side (Env-side) RA deployments and identify five fundamental EM reconfiguration dimensions: position, orientation, radiation pattern, polarization, and frequency response. Building on this framework, we examine how these dimensions customize communication channels, actively control target information mapping for sensing, and reshape the interplay between S&C, particularly by improving compatibility between S&C requirements under shared resources. Two case studies illustrate how TSc-side and Env-side EM reconfiguration can improve spatial selectivity and the achievable S&C performance tradeoff. Finally, we discuss several promising directions for advancing RA-enhanced ISAC systems, including fundamental performance limits, low-overhead black-box optimization, and artificial intelligence (AI)-enabled semantic ISAC.

eess.SP↗

arXiv2026

Supporting Habit Formation in Pre-Meal Photo Capture: A Longitudinal Study of Behavior Change Techniques

Pre-meal photo capture is lightweight only with a smartphone, yet habit formation requires repeatedly initiating and completing the action within the narrow window before the first bite. We examine how supports based on behavior change techniques can shape capture adherence, perceived habit, and support needs over time. Using a participant-informed selection process, we implemented three smartphone supports targeting Prompts/cues, Focus on past success, and Social support. We evaluated them in a five-week field study with 54 participants, spanning support, withdrawal, and reinstatement, followed by interviews and a three-week follow-up. Across 4,945 main meals, 63.9\% were confirmed pre-meal captures, and adherence was broadly maintained across phases. Perceived automaticity and spontaneous remembering improved mainly during initial support. As capture became more familiar, participants combined spontaneous remembering with deliberate strategies, reduced active feature use while continuing to value support availability, and carried the practice forward through accumulated records, familiar tools, and chosen relationships. These findings frame BCT support for pre-meal photo capture as a configurable arrangement. Systems should preserve a predictable capture pathway while adapting when, how, and through whom support remains available.

cs.HC↗

arXiv2026

A GeV PWN Candidate in the PSR J1135--6055/G293.8+0.6 Field

We present a \textit{Fermi} Large Area Telescope (LAT) study of PSR~J1135--6055 and the composite supernova remnant (SNR) G293.8+0.6. MeerKAT radio observations reveal a bright central plerion within the faint SNR shell, while Chandra observations detect a compact nebula around the pulsar. Together, these observations establish the presence of a PWN within G293.8+0.6. Our analysis of the 20~GeV--1~TeV LAT data yields a point-like excess with $\mathrm{TS}=26.30$ and photon index $Γ=1.63\pm0.33$. Its best-fit position is consistent with the pulsar within the 95\% statistical localization region, and neither the uniform-disk nor Gaussian model is favored, providing no evidence for significant extension. A one-zone leptonic model, together with an independent jet component, reproduces the multiwavelength observations. In this model, the GeV photons are produced mainly by inverse-Compton scattering from relativistic electrons and positrons. Given its positional agreement with the pulsar/PWN and broadband interpretation, we regard the LAT point source as a GeV PWN candidate.

astro-ph.HE↗

arXiv2026

Heterogeneous Co-Evolution Learning for Tool-Efficient Search Agents with Zero Labels

Deep search agents extend language models by interleaving reasoning with external retrieval. However, effective search-agent training requires addressing two challenges beyond final-answer optimization. First, the capability frontier of agents is largely determined by training questions, whereas fixed or labeled task collections cannot adapt as agents improve. Second, outcome-only reinforcement learning treats all successful trajectories equally, even when some correct answers rely on redundant searches, creating an inherent tension between answer accuracy and tool efficiency. To address these issues, we introduce \textsc{ZEST-Search}, a zero-label and tool-efficient framework for search-agent training via heterogeneous co-evolution. Starting from unlabeled reference questions, \textsc{ZEST-Search} constructs validated search tasks online through an answer-first and evidence-path-guided generation process. A heterogeneous proposer adaptively constructs training tasks around the solver's evolving capability boundary, while the solver learns search-augmented reasoning through reinforcement learning. To jointly optimize task difficulty and search efficiency, \textsc{ZEST-Search} employs an intermediate-difficulty reward for the proposer and a task-local tool-aware reward for the solver, encouraging informative tasks and correct answers with fewer redundant searches. Experiments on seven general-domain and multi-hop question answering benchmarks demonstrate that \textsc{ZEST-Search} consistently improves search-agent performance while achieving higher tool efficiency without labeled search-training data or expert search trajectories. Further analyses show that heterogeneous generation, adaptive difficulty control, and tool-aware optimization enable more effective search-agent training.

cs.LG↗

arXiv2026

Exploiting Beam Squint for Near-Field Wideband Beamforming in Multi-User OFDMA Systems

The beam squint effect has traditionally been viewed as detrimental to near-field wideband beamforming due to the reduction of beamforming gain towards the desired direction. In this paper, we show that it can instead be harnessed as a benefit in a hardware-constrained orthogonal frequency division multiple access (OFDMA) system, where the system, equipped with a single radio-frequency (RF) chain, aims to serve multiple spatially distributed users by allocating different subcarriers to different users. By exploiting the beam squint effect, flexible frequency-dependent beamfocusing patterns are realized via parallel true-time-delay (TTD) units, enabling nearly full beamforming gain across users. In particular, we formulate a max-min rate optimization problem that jointly designs the time delays, subcarrier-user association, analog beamforming, and power allocation. To tackle the resulting nonconvex mixed-integer problem, a two-stage optimization framework is proposed. In Stage~I, subcarrier-user association and structured analog beamforming are jointly optimized. In Stage~II, power allocation is obtained via bisection search and inverse water-filling. Simulation results demonstrate that the proposed method can achieve a significant improvement by utilizing beam squint to serve multiple spatially distributed users in OFDMA systems.

eess.SP↗

arXiv2026

Enabling Preference-driven Unlearning in Few-step Distilled Text-to-Image Diffusion Models

Text-to-image diffusion models are increasingly distilled into few-step variants and being deployed to enable fast inference. However, their ability to generate harmful or undesired content poses significant safety risks. Data-driven unlearning methods suppress targeted generations by fine-tuning model weights using specialized unlearning objectives. Crucially, these objectives implicitly rely on multi-step denoising dynamics, an assumption that breaks down for few-step distilled (FSD) models, resulting in ineffective forgetting. Furthermore, performing unlearning on the non-distilled base model and subsequently re-distilling it to obtain an unlearned FSD model incurs substantial computational and time overhead, making it impractical in many settings. Hence, we address this limitation with a preference-driven unlearning framework that revisits Direct Preference Optimization (DPO) for diffusion models. We show that standard DPO and its unlearning derivatives, formulated around noise-prediction error, transfer poorly to FSD models due to their altered generation dynamics. To overcome this, we introduce a modified preference optimization formulation explicitly aligned with the few-step generation properties, enabling direct concept removal in FSD models while preserving few-step efficiency and maintaining strong retention of desirable (non-targeted) capabilities. We evaluate our framework primarily on identity and NSFW (nudity) removal tasks and also extend our method to object-level unlearning. Extensive experiments demonstrate consistent and effective forgetting, and strong retention performance, establishing our method as a practical and principled solution for unlearning in FSD models.

cs.CV↗

arXiv2026

StoryEcho: A Narrative Mirroring Loop Generative Storytelling System for Picky-Eating Intervention

Picky eating can limit children's dietary variety and create tension in family feeding routines. Existing food-related technologies often focus on mealtime intervention or standalone educational artifacts, offering limited support for connecting low-pressure narrative engagement with children's real-world food exploration over time. We present StoryEcho, a generative storytelling system centered on a narrative mirroring loop, in which personalized stories model sensory exploration through a persistent counterpart and children's subsequent food encounters are reflected back into narrative feedback and future story development. Informed by a formative study, we designed StoryEcho and evaluated it in a 14-day between-subjects field study with 26 families. Compared with food-personalized generative stories without the narrative mirroring loop, StoryEcho was associated with higher try level, approach, and intake, and lower resistance and caregiver pressure. These findings suggest that narrative mirroring can support children's low-pressure food exploration in family routines, while highlighting design tensions for future generative storytelling interventions.

cs.HC↗

arXiv2026

Reconstruction and Reflection of Positive Experiences through Resurfacing Laughter-indexed Everyday Moments

Positive everyday moments often escape deliberate recording, while continuous self-tracking can generate extensive records that are difficult to revisit. We explore laughter as a naturally occurring, sparse index for constructing contextualized personal records to support later reconstruction and reflection. A formative study with 12 participants characterized laughter as an affective but semantically incomplete index and informed \textit{LaughAnchor}, a mobile and wearable self-tracking system. During participant-initiated recording, the system assembles detected laughter and aligned context into candidate moments for later reconstruction and reflection, with layered context disclosure, user-controlled curation, and near-term and long-term resurfacing. In a three-week field deployment with 12 participants, passive indexing preserved moments they considered unlikely to record deliberately but valued retrospectively. During resurfacing, participants attributed affective re-experiencing to laughter and used additional context both to reconstruct episodes and to explore already-recalled experiences. Across moments and reviews, resurfacing supported rediscovery and broader awareness of relationships, routines, and emotional states. These findings inform self-tracking designs that use sparse affective indices to organize contextual records for reconstruction and reflection, while keeping interpretation and retention under user control.

cs.HC↗

arXiv2026

Earinter: A Closed-Loop System for Eating Pace Regulation with Just-in-Time Intervention Using Commodity Earbuds

Rapid eating is common yet difficult to regulate in situ, partly because people seldom notice pace changes and sustained self-monitoring is effortful. We present Earinter, a commodity-earbud-based closed-loop system that integrates in-the-wild sensing, real-time reasoning, and theory-grounded just-in-time (JIT) intervention to regulate eating pace during daily meals. Earinter repurposes the earbud's bone-conduction voice sensor to capture chewing-related vibrations and estimate eating pace as chews per swallow (CPS) for on-device inference. With data collected equally across \modify{sessions in quiet and noisy environments}, Earinter achieves reliable chewing detection (F1 = 0.97) and accurate eating pace estimation (MAE: 0.18 $\pm$ 0.13 chews/min, 3.65 $\pm$ 3.86 chews/swallow), enabling robust tracking for closed-loop use. Guided by Dual Systems Theory and refined through two Wizard-of-Oz pilots, Earinter adopts a user-friendly design for JIT intervention content and delivery policy in daily meals. In a 13-day within-subject field study (N=14), the closed-loop system significantly increased CPS and slowed meal-level eating pace, with statistical signs of carryover on retention-probe days and generally favorable comfort and usability. These effects should be interpreted as changes in eating-pace regulation rather than evidence of weight loss. Our findings highlight how single-modality commodity earables can support practical, theory-driven closed-loop JIT interventions for regulating eating pace in the wild.

cs.HC↗

arXiv2026

Phoenix-VAD: Streaming Semantic Endpoint Detection for Full-Duplex Speech Interaction

Spoken dialogue models have significantly advanced intelligent human-computer interaction, yet they lack a plug-and-play full-duplex prediction module for semantic endpoint detection, hindering seamless audio interactions. In this paper, we introduce Phoenix-VAD, an LLM-based model that enables streaming semantic endpoint detection. Specifically, Phoenix-VAD leverages the semantic comprehension capability of the LLM and a sliding window training strategy to achieve reliable semantic endpoint detection while supporting streaming inference. Experiments on both semantically complete and incomplete speech scenarios indicate that Phoenix-VAD achieves excellent and competitive performance. Furthermore, this design enables the full-duplex prediction module to be optimized independently of the dialogue model, providing more reliable and flexible support for next-generation human-computer interaction.

eess.AS↗

arXiv2026

Super-Resolution Initialization of High-Fidelity CFD Simulations for Pebble-Bed Reactors

High-order CFD simulations provide detailed resolution of the heterogeneous interstitial flow in pebble-bed reactors, but their computational cost is high, especially during the initial flow-development period required to reach statistically stationary conditions. This work investigates the use of a Super-Resolution Graph Neural Network (SR-GNN) to improve the initialization of high-order NekRS simulations. Lower-order P = 2 velocity fields are used as inputs to reconstruct higher-order representations, which are then used as initial conditions for P = 7 restart simulations. The approach is evaluated using a 146-pebble bed at Re = 1000, Re = 2500, and Re = 5000, with pressure-drop convergence used as the main figure of merit. The SR-GNN models were trained using paired low- and high-order snapshots and were first evaluated through qualitative inference comparisons. High-order restart simulations showed that, for Re = 1000 and Re = 2500, the SR-GNN initialized cases produced pressure-drop histories similar to direct restarts from true P = 2 fields. For Re = 5000, however, the super-resolved field restart approached the statistically stationary P = 7 pressure-drop range faster than both the direct P = 2 restart and the reference P = 7 simulation initialized from a uniform velocity field. The trained Re = 5000 model was also applied to a larger 1568-pebble bed, demonstrating qualitative applicability of the workflow to a significantly larger packed-bed geometry. These results indicate that SR-GNN-based initialization is a promising strategy for reducing high-order flow-development cost, while also motivating further work on broader Reynolds-number and geometry generalization.

physics.flu-dyn↗

arXiv2026

Rethinking Normalization Placement for LLMs: Post-Norm under Curriculum Depth Growing

Pre-norm is the standard normalization placement in modern Transformers because it facilitates joint optimization of full-depth models. We ask whether this preference persists when depth is introduced through a curriculum. In curriculum depth growth, each appended block receives the boundary representation produced by a trained prefix, making normalization placement relevant to forward conditioning. We therefore test whether placement and training curriculum interact. In a controlled distillation study with a Qwen3-8B teacher and a nine-layer student, pre-norm and post-norm are indistinguishable under joint training, differing by $0.0004$ validation CE, while post-norm improves over pre-norm by $0.0328$ under curriculum growth, an order of magnitude larger. A post-joint control matched by student active-layer tokens remains worse than post-grow, which rules out compute as the sole explanation. The ranking crosses over during the curriculum: post-norm takes the lead once blocks are appended. Single-block and freeze controls localize the ranking change to block appending rather than shallow-block quality or retraining. Boundary diagnostics associate post-norm with stable residual scales and pre-norm with structural-token scale drift; on a fixed batch, the final pre-grow block is also nearly identity-mapped. Together with the phase-wise crossover, these observations are consistent with boundary-scale conditioning after new blocks are appended. The results motivate treating normalization placement and training curriculum as coupled design choices in this distillation setting.

cs.AI↗

arXiv2026

Phoenix TTS: High-Fidelity Synthesis and Voice Conversion via Flow-Matching-Driven Speech Tokenization

In current zero-shot text-to-speech systems, conventional semantic tokenizers are typically optimized using supervised automatic speech recognition or self-supervised learning objectives. However, due to the inherent nature of speech, semantic and acoustic information cannot be completely decoupled, and ASR-based tokenizers discard acoustic details to focus on linguistic content; models relying on them usually struggle to achieve optimal speaker similarity. Furthermore, these tokenizers are optimized independently and lack direct supervision from downstream acoustic generation tasks. This isolated training creates a feature gap between the extracted discrete tokens and the continuous space required by acoustic models, fundamentally bottlenecking the upper bound of synthesis quality. To bridge this gap, we propose Phoenix TTS, a unified framework that tightly couples representation learning with generative acoustic modeling. Specifically, our speech tokenizer is optimized to reconstruct self-supervised features to maintain semantic richness, while simultaneously receiving direct supervision from a Flow Matching training loss. Through this joint training paradigm, the extracted discrete tokens successfully preserve essential semantic information and natively align with the feature space of the downstream Flow Matching model. Comprehensive evaluations highlight the efficiency and effectiveness of Phoenix TTS. Trained on 110K hours of data, the system achieves excellent speech intelligibility, yielding WER that consistently falls below that of ground-truth recordings. Simultaneously, it maintains robust zero-shot speaker similarity, rivaling or outperforming several prominent large-scale baselines. Furthermore, as an advantageous byproduct of this unified training, the learned tokenizer can be seamlessly adapted to zero-shot voice conversion tasks without requiring task-specific fine-tuning.

cs.SD↗

arXiv2026

MMAC: A Massive Multi-dimensional Benchmark for Audio Captioning

With the development of audio large language models (AudioLLMs), audio captioning needs to move from brief descriptions toward open-ended and fine-grained free-form descriptions. Existing evaluations often focus on generation quality or task performance, making it difficult to diagnose information coverage and description reliability. We propose MMAC, a \textbf{M}assive \textbf{M}ulti-dimensional benchmark for \textbf{A}udio \textbf{C}aptioning. MMAC contains 5,638 audio clips from more than 20 data sources, covering 6 capability categories and 15 evaluation dimensions. Given a model-generated caption, MMAC checks whether it mentions relevant information in the target dimension and whether the mentioned content is consistent with the reference label. We evaluate representative open-source and proprietary AudioLLMs. Results show clear differences across evaluation dimensions, information coverage, and description reliability. We will release the MMAC benchmark and evaluation code.

cs.SD↗

arXiv2026

Zeroth-Order Blind Interference Suppression for Multi-RIS-Aided Wireless Systems

In this paper, we study measurement-driven signal-to-interference-plus-noise ratio (SINR) maximization for a multi-reconfigurable intelligent surface (RIS)-aided single-input single-output (SISO) system with unknown strong interference sources. Specifically, the objective is to optimize the reflection coefficients such that the SINR is maximized at the receiver. As the interference channels are unknown, such an optimization problem is a black-box optimization problem with an objective function whose closed-form analytical expression is unknown. To address the high-dimensional black-box optimization problem with discrete variable constraints, we introduce a group-based phase parameterization that significantly reduces the search dimension. Building on this model, we develop a group-based zeroth-order adaptive moment (ZO-AdaMM) algorithm. Simulation results show that the proposed grouping strategy markedly accelerates the convergence speed and achieves a superior interference suppression performance under limited measurement budgets, especially in the small-budget regime.

eess.SP↗

arXiv2026

Integrated Channel Estimation and Sensing for Near-Field ELAA Systems via Low-Rank Tensor Decomposition

In this paper, we study the problem of uplink channel estimation for near-filed orthogonal frequency division multiplexing (OFDM) systems, where a base station (BS), equipped with an extremely large-scale antenna array (ELAA), serves multiple users over the same time-frequency resource block. A non-orthogonal pilot transmission scheme is considered to accommodate a larger number of users that can be supported by ELAA systems without incurring an excessive amount of training overhead. To facilitate efficient multi-user channel estimation, we express the received signal as a third-order low-rank tensor, which admits a canonical polyadic decomposition (CPD) model for line-of-sight (LoS) scenarios and a block term decomposition (BTD) model for non-line-of-sight (NLoS) scenarios. An alternating least squares (ALS) algorithm and a non-linear least squares (NLS) algorithm are employed to perform CPD and BTD, respectively. Channel parameters are then efficiently extracted from the recovered factor matrices. By exploiting the geometry of the propagation paths in the estimated channel, users' positions can be precisely determined in LoS scenarios. Moreover, our uniqueness analysis shows that the proposed tensor-based joint multi-user channel estimation framework is effective even when the number of pilot symbols is much smaller than the number of users, revealing its potential in training overhead reduction. Simulation results demonstrate that the proposed method achieves markedly higher channel estimation accuracy than compressed sensing (CS)-based approaches.

eess.SP↗

arXiv2026

DeepTravel: An End-to-End Agentic Reinforcement Learning Framework for Autonomous Travel Planning Agents

Travel planning (TP) agent has recently worked as an emerging building block to interact with external tools/resources for travel itinerary generation, ensuring an enjoyable user experience. Despite its benefits, existing studies rely on hand-craft prompt and fixed agent workflow, hindering more flexible and autonomous TP agents. This paper proposes DeepTravel, an end-to-end agentic reinforcement learning framework for building an autonomous travel planning agent, capable of autonomously planning, executing tools, and reflecting on tool responses to explore, verify, and refine intermediate actions in multi-step reasoning. To achieve this, we first construct a robust travel sandbox by caching transportation, accommodation and POI data, facilitating TP agent training without being constrained by real-world APIs limitations (e.g., inconsistent outputs). Moreover, we develop a hierarchical reward modeling system, where a trajectory-level verifier first checks spatiotemporal feasibility and filters unsatisfied travel itinerary, and then the turn-level verifier further validate itinerary's detail consistency with tool responses, enabling efficient and precise reward service. Finally, we propose the reply-augmented reinforcement learning method that enables TP agent to periodically replay from a failure experience buffer, emerging notable agentic capacity. We deploy the trained TP agent in the DiDi Enterprise Solutions application. A three-month online test shows that it achieves 82% accuracy in travel itinerary generation. Comprehensive offline evaluations further demonstrate that DeepTravel enables small-sized LLMs (e.g., Qwen3-32B) to significantly outperform frontier LLMs (e.g., OpenAI o1/o3 and DeepSeek-R1) and existing TP agent frameworks.

cs.AI↗

arXiv2026

JD Oxygen AI Item Center (Oxygen AIIC) V1: An Industrial-Scale LLM/VLM-Centric Solution for Item Understanding, Management, and Applications

JD$.$com, one of the world's largest e-commerce platforms, serves over 700 million active users and millions of merchants, with a catalog of tens of billions of SKUs. At this scale, high-quality, structured item knowledge underpins a better consumer experience, lower management costs, and higher operational efficiency-yet producing and serving it poses three industrial-scale challenges: fast-emerging concepts, high-quality knowledge production for massive SKUs, and diverse downstream requirements. To address these challenges, we present the JD Oxygen AI Item Center (Oxygen AIIC), an industrial-scale platform built on LLMs/VLMs for item-knowledge production and service. Oxygen AIIC is built around four core pillars: (i) ontology engineering driven by efficient human-AI collaboration, which supports the dynamic evolution and agile expansion of an ontology with millions of entries; (ii) a "Semantic Search then Discrimination"(S2D) knowledge identification architecture that, combined with throughput improvement strategies, enables scalable, extensible, and high-throughput AI Item Library production for tens of billions of SKUs; (iii) self-evolving item-understanding LLMs/VLMs that improve in a stable and controllable manner, enabling knowledge production with 94.2% precision and 82.8% recall; and (iv) a unified item tunnel that serves as the data and service hub. Oxygen AIIC now covers tens of thousands of JD categories and processes hundreds of millions of item updates per day on Huawei Ascend NPUs. It has accumulated hundreds of billions of item-knowledge assets. Deployed across core business scenarios-including search, recommendation, operations, category planning-Oxygen AIIC has delivered measurable gains at scale. Search-traffic coverage reaches 80.4%, item-information quality issues drop by 37%, the automated fill rate of core attributes during item listing exceeds 80%.

cs.AI↗