arXiv:2602.13640v1 Announce Type: cross Abstract: Existing robotic manipulation methods primarily rely on visual and proprioceptive observations, which may struggle to infer contact-related interaction states in partially observable real-world environments. Acoustic cues, by contrast, naturally encode rich interaction dynamics during contact, yet remain underexploited in current multimodal fusion literature. Most multimodal fusion approaches implicitly assume homogeneous roles across modalities, and thus design flat and symmetric fusion structures. However, this assumption is ill-suited for acoustic signals, which are inherently sparse and contact-driven. To achieve precise robotic manipulation through acoustic-informed perception, we propose a hierarchical representation fusion framework that progressively integrates audio, vision, and proprioception. Our approach first conditions visual and proprioceptive representations on acoustic cues, and then explicitly models higher-order cross-modal interactions to capture complementary dependencies among modalities. The fused representation is leveraged by a diffusion-based policy to directly generate continuous robot actions from multimodal observations. The combination of end-to-end learning and hierarchical fusion structure enables the policy to exploit task-relevant acoustic information while mitigating interference from less informative modalities. The proposed method has been evaluated on real-world robotic manipulation tasks, including liquid pouring and cabinet opening. Extensive experiment results demonstrate that our approach consistently outperforms state-of-the-art multimodal fusion frameworks, particularly in scenarios where acoustic cues provide task-relevant information not readily available from visual observations alone. Furthermore, a mutual information analysis is conducted to interpret the effect of audio cues in robotic manipulation via multimodal fusion.

arXiv:2602.14301v1 Announce Type: cross Abstract: Recent Mixture-of-Experts (MoE)-based large language models (LLMs) such as Qwen-MoE and DeepSeek-MoE are transforming generative AI in natural language processing. However, these models require vast and diverse training data. Federated learning (FL) addresses this challenge by leveraging private data from heterogeneous edge devices for privacy-preserving MoE training. Nonetheless, traditional FL approaches require devices to host local MoE models, which is impractical for resource-constrained devices due to large model sizes. To address this, we propose DeepFusion, the first scalable federated MoE training framework that enables the fusion of heterogeneous on-device LLM knowledge via federated knowledge distillation, yielding a knowledge-abundant global MoE model. Specifically, DeepFusion features each device to independently configure and train an on-device LLM tailored to its own needs and hardware limitations. Furthermore, we propose a novel View-Aligned Attention (VAA) module that integrates multi-stage feature representations from the global MoE model to construct a predictive perspective aligned with on-device LLMs, thereby enabling effective cross-architecture knowledge distillation. By explicitly aligning predictive perspectives, VAA resolves the view-mismatch problem in traditional federated knowledge distillation, which arises from heterogeneity in model architectures and prediction behaviors between on-device LLMs and the global MoE model. Experiments with industry-level MoE models (Qwen-MoE and DeepSeek-MoE) and real-world datasets (medical and finance) demonstrate that DeepFusion achieves performance close to centralized MoE training. Compared with key federated MoE baselines, DeepFusion reduces communication costs by up to 71% and improves token perplexity by up to 5.28%.

arXiv:2602.14462v1 Announce Type: cross Abstract: Data-parallel (DP) training with synchronous all-reduce is a dominant paradigm for full-parameter fine-tuning of large language models (LLMs). While parameter synchronization guarantees numerical equivalence of model weights after each iteration, it does not necessarily imply alignment of worker-level optimization dynamics before gradient aggregation. This paper identifies and studies this latent mismatch, termed \emph{silent inconsistency}, where cross-worker divergence in losses and gradients can remain invisible under conventional aggregated monitoring signals. We propose a lightweight, model-agnostic diagnostic framework that quantifies worker-level consistency using training signals readily available in standard pipelines. Specifically, we introduce three complementary metrics: loss dispersion, gradient-norm dispersion, and gradient-direction consistency measured by inter-worker cosine similarity. The proposed metrics incur negligible overhead and require no modification to model architecture, synchronization mechanisms, or optimization algorithms. We validate the framework by fully fine-tuning the 1B-parameter \texttt{openPangu-Embedded-1B-V1.1} model on the \texttt{tatsu-lab/alpaca} dataset using an 8-NPU DP setup, under controlled perturbations of cross-rank stochasticity. Experimental results show that progressively desynchronized data shuffling and random seeds lead to substantial increases in loss/gradient dispersion and reduced directional alignment, despite smooth globally averaged loss curves. These findings demonstrate that the proposed indicators provide actionable visibility into hidden instability modes in large-scale DP fine-tuning, enabling more reliable diagnosis and configuration assessment.

arXiv:2602.14681v1 Announce Type: cross Abstract: LLM-powered Multi-Agent Systems (MAS) have emerged as an effective approach towards collaborative intelligence, and have attracted wide research interests. Among them, ``self-evolving'' MAS, treated as a more flexible and powerful technical route, can construct task-adaptive workflows or communication topologies, instead of relying on a predefined static structue template. Current self-evolving MAS mainly focus on Spatial Evolving or Temporal Evolving paradigm, which only considers the single dimension of evolution and does not fully incentivize LLMs' collaborative capability. In this work, we start from a novel Spatio-Temporal perspective by proposing ST-EVO, which supports dialogue-wise communication scheduling with a compact yet powerful flow-matching based Scheduler. To make precise Spatio-Temporal scheduling, ST-EVO can also perceive the uncertainty of MAS, and possesses self-feedback ability to learn from accumulated experience. Extensive experiments on nine benchmarks demonstrate the state-of-the-art performance of ST-EVO, achieving about 5%--25% accuracy improvement.

arXiv:2512.13697v2 Announce Type: replace-cross Abstract: The proliferation of Large Language Models (LLMs) raises a critical question about what it means to be human when we share an increasingly symbiotic relationship with persuasive and creative machines. This paper examines patterns of human-AI coevolution in creative writing, investigating how human craft and agency are adapting alongside machine capabilities. We challenge the prevailing notion of stylistic homogenization by examining diverse patterns in longitudinal writing data. Using a large-scale corpus spanning the pre- and post-LLM era, we observe patterns suggestive of a "Dual-Track Evolution": thematic convergence around AI-related topics, coupled with structured stylistic differentiation. Our analysis reveals three emergent adaptation patterns: authors showing increased similarity to AI style, those exhibiting decreased similarity, and those maintaining stylistic stability while engaging with AI-related themes. This Creative Archetype Map illuminates how authorship is coevolving with AI, contributing to discussions about human-AI collaboration, detection challenges, and the preservation of creative diversity.

arXiv:2602.11858v2 Announce Type: replace-cross Abstract: Multimodal Large Language Models (MLLMs) excel at broad visual understanding but still struggle with fine-grained perception, where decisive evidence is small and easily overwhelmed by global context. Recent "Thinking-with-Images" methods alleviate this by iteratively zooming in and out regions of interest during inference, but incur high latency due to repeated tool calls and visual re-encoding. To address this, we propose Region-to-Image Distillation, which transforms zooming from an inference-time tool into a training-time primitive, thereby internalizing the benefits of agentic zooming into a single forward pass of an MLLM. In particular, we first zoom in to micro-cropped regions to let strong teacher models generate high-quality VQA data, and then distill this region-grounded supervision back to the full image. After training on such data, the smaller student model improves "single-glance" fine-grained perception without tool use. To rigorously evaluate this capability, we further present ZoomBench, a hybrid-annotated benchmark of 845 VQA data spanning six fine-grained perceptual dimensions, together with a dual-view protocol that quantifies the global--regional "zooming gap". Experiments show that our models achieve leading performance across multiple fine-grained perception benchmarks, and also improve general multimodal cognition on benchmarks such as visual reasoning and GUI agents. We further discuss when "Thinking-with-Images" is necessary versus when its gains can be distilled into a single forward pass. Our code is available at https://github.com/inclusionAI/Zooming-without-Zooming.