Background: As generative AI (GenAI) becomes increasingly prevalent, its impact on physician mental health has garnered significant attention; yet, empirical evidence remains limited. Objective: This study aims to investigate the correlations between the usage frequency of GenAI, perceived usefulness (PU), and perceived risk (PR) of GenAI with physicians’ burnout and professional fulfillment. Methods: A mixed methods design was used, integrating a quantitative survey of physicians across 4 regions in China with in-depth qualitative interviews to elucidate the underlying psychological mechanisms. The quantitative component involved a cross-sectional survey of 961 physicians, with the questionnaire collecting data on demographic and professional characteristics, socioeconomic status, GenAI usage frequency, PU, and PR. Semistructured interviews with 10 physicians were used for in-depth mining. Multivariable logistic and linear regression models with province-level fixed effects were fitted to examine the association between usage of GenAI, PU, PR, and physicians’ burnout and fulfillment. Stratified analyses were further performed to explore the moderating effect of demographic and clinical characteristics. Results: Quantitative analysis revealed no direct correlation between GenAI usage frequency and burnout. However, PU was positively associated with professional fulfillment (odds ratio [OR] 1.56, 95% CI 1.17-2.08; P=.003), whereas PR was associated with a higher likelihood of burnout (OR 1.80, 95% CI 1.46-2.21; P<.001). Stratified analyses showed that for physicians working ≥3 night shifts per week, GenAI usage was associated with higher odds of burnout, although the estimate was imprecise (OR 13.96, 95% CI 2.40-81.04; P=.003). The qualitative findings further suggested that the benefits of using GenAI may be offset by the additional burden. The PU of GenAI was perceived to enhance professional fulfillment by bolstering self-efficacy, whereas the PR of GenAI was linked to heightened burnout rooted in unclear boundaries of responsibilities and rights, as well as challenges to professional identity. Conclusions: The GenAI revolution in medicine is as much a psychological transition as it is a technological one. GenAI use is not directly associated with improved psychological states among clinicians. The PU of GenAI relates to professional fulfillment, and the PR concerns correspond to elevated burnout. Sustaining clinician well-being during this digital shift thus parallels a dual requirement, balancing the potential for professional fulfillment tied to GenAI utility against the concurrent verification fatigue and legal uncertainty cluster around clinician burnout.

arXiv:2609.10372v1 Announce Type: cross Abstract: We present the PACE, a framework for retrieval-augmented dialogue serving that formalizes Perceived Time-to-First-Response (PTFR) as a QoE objective and minimizes it under quality/cost constraints. Unlike prior work on cascaded routing, semantic caching, or adaptive retrieval, PACE jointly controls which answer source composes the response and what fills the waiting window. Deployed on a humanoid-robot sales service, it combines three mechanisms: a load-adaptive cascading router, a joint path-filler controller, and volatility-aware cache admission. On 75k CarQA requests, the cascade halves pure-LLM PTFR at P95 (0.29 vs 0.53s at c16). The adaptive controller reaches 0.41s P95, outperforming RAG by 2.4 times at high load with equal quality. The filler controller cuts calls by 94% with zero conflict. Volatility-aware admission reduces stale answers from 86% to 0%. A gating rule ensures the controller never worse than the baseline, with exposure bounded by one hold period. This is the first quantification of filler-answer conflict risk in deployed services.

arXiv:2605.25422v1 Announce Type: cross Abstract: The convergence of large language models (LLMs) with 6G networks is fostering a paradigm of autonomous multi-agent cooperation, which in turn is expected to substantially increase east-west traffic. Although latent-space interaction mechanisms can enable more efficient collaboration than symbolic natural-language (NL) exchanges, prior work often abstracts away the associated communication overhead under practical wireless constraints. In embodied multi-agent settings, heterogeneous interaction media incur disparate inference and transmission costs, thereby inducing an inherent end-to-end (E2E) latency trade-off. To address this, we propose a joint design that integrates communication-media selection with wireless resource allocation. Through analytical characterization and simulation-based evaluation, we show that neither token-based transmission nor key-value (KV) cache-based transmission is uniformly optimal across operating regimes, as performance depends critically on system parameters such as available computational resources and channel conditions. Accordingly, we formulate a joint optimization problem aimed at minimizing the E2E latency of multi-agent collaboration and develop a low-complexity joint media selection and resource allocation (JMSRA) algorithm. Numerical results further confirm that, by adaptively coordinating the interaction media and bandwidth allocation over heterogeneous links, the proposed scheme achieves markedly reduced E2E latency relative to conventional NL-only and KV-cache-only baselines, enabling efficient and robust multi-agent collaboration in future wireless networks.

arXiv:2603.19609v2 Announce Type: replace-cross Abstract: We present LoD-Loc v3, a novel method for generalized aerial visual localization in dense urban environments. While prior work LoD-Loc v2 achieves localization through semantic building silhouette alignment with low-detail city models, it suffers from two key limitations: poor cross-scene generalization and frequent failure in dense building scenes. Our method addresses these challenges through two key innovations. First, we develop a new synthetic data generation pipeline that produces InsLoD-Loc - the largest instance segmentation dataset for aerial imagery to date, comprising 100k images with precise instance building annotations. This enables trained models to exhibit remarkable zero-shot generalization capability. Second, we reformulate the localization paradigm by shifting from semantic to instance silhouette alignment, which significantly reduces pose estimation ambiguity in dense scenes. Extensive experiments demonstrate that LoD-Loc v3 outperforms existing state-of-the-art (SOTA) baselines, achieving superior performance in both cross-scene and dense urban scenarios with a large margin. The project is available at https://nudt-sawlab.github.io/LoD-Locv3/.

arXiv:2603.03825v1 Announce Type: cross Abstract: The cold-start initialization stage plays a pivotal role in training Multimodal Large Reasoning Models (MLRMs), yet its mechanisms remain insufficiently understood. To analyze this stage, we introduce the Visual Attention Score (VAS), an attention-based metric that quantifies how much a model attends to visual tokens. We find that reasoning performance is strongly correlated with VAS (r=0.9616): models with higher VAS achieve substantially stronger multimodal reasoning. Surprisingly, multimodal cold-start fails to elevate VAS, resulting in attention distributions close to the base model, whereas text-only cold-start leads to a clear increase. We term this counter-intuitive phenomenon Lazy Attention Localization. To validate its causal role, we design training-free interventions that directly modulate attention allocation during inference, performance gains of 1$-$2% without any retraining. Building on these insights, we further propose Attention-Guided Visual Anchoring and Reflection (AVAR), a comprehensive cold-start framework that integrates visual-anchored data synthesis, attention-guided objectives, and visual-anchored reward shaping. Applied to Qwen2.5-VL-7B, AVAR achieves an average gain of 7.0% across 7 multimodal reasoning benchmarks. Ablation studies further confirm that each component of AVAR contributes step-wise to the overall gains. The code, data, and models are available at https://github.com/lrlbbzl/Qwen-AVAR.