PSGS: Text-driven Panorama Sliding Scene Generation via Gaussian Splatting

Generating realistic 3D scenes from text is crucial for immersive applications like VR, AR, and gaming. While text-driven approaches promise efficiency, existing methods suffer from limited 3D-text data and inconsistent multi-view stitching, resulting in overly simplistic scenes. To address this, we propose PSGS, a two-stage framework for high-fidelity panoramic scene generation. First, a novel two-layer optimization architecture generates semantically coherent panoramas: a layout reasoning layer parses text into structured spatial relationships, while a self-optimization layer refines visual details via iterative MLLM feedback. Second, our panorama sliding mechanism initializes globally consistent 3D Gaussian Splatting point clouds by strategically sampling overlapping perspectives. By incorporating depth and semantic coherence losses during training, we greatly improve the quality and detail fidelity of rendered scenes. Our experiments demonstrate that PSGS outperforms existing methods in panorama generation and produces more appealing 3D scenes, offering a robust solution for scalable immersive content creation.

RASST: Fast Cross-modal Retrieval-Augmented Simultaneous Speech Translation

Simultaneous speech translation (SST) produces target text incrementally from partial speech input. Recent speech large language models (Speech LLMs) have substantially improved SST quality, yet they still struggle to correctly translate rare and domain-specific terminology. While retrieval augmentation has been effective for terminology translation in machine translation, bringing retrieval to SST is non-trivial: it requires fast and accurate cross-modal (speech-to-text) retrieval under partial, continually arriving input, and the model must decide whether and when to apply retrieved terms during incremental generation. We propose Retrieval-Augmented Simultaneous Speech Translation (RASST), which tightly integrates cross-modal retrieval into the SST pipeline. RASST trains a lightweight speech-text retriever and performs efficient sliding-window retrieval, providing chunkwise terminology hints to the Speech LLM. We further synthesize training data that teaches the Speech LLM to leverage retrieved terms precisely. Experiments on three language directions of the ACL 60/60 dev set show that RASST improves terminology translation accuracy by up to 16% and increases overall translation quality by up to 3 BLEU points, with ablations confirming the contribution of each component.

PEMNet: Towards Autonomous and Enhanced Environment-Aware Mobile Networks

With 5G deployment and the evolution toward 6G, mobile networks must make decisions in highly dynamic environments under strict latency, energy, and spectrum constraints. Achieving this goal, however, depends on prior knowledge of spatial-temporal variations in wireless channels and traffic demands. This motivates a joint, site-specific representation of radio propagation and user demand that is queryable at low online overhead. In this work, we propose the perception embedding map (PEM), a localized framework that embeds fine-grained channel statistics together with grid-level spatial-temporal traffic patterns over a base station's coverage. PEM is built from standard-compliant measurements -- such as measurement report and scheduling/quality-of-service logs -- so it can be deployed and maintained at scale with low cost. Integrated into PEM, this joint knowledge supports enhanced environment-aware optimization across PHY, MAC, and network layers while substantially reducing training overhead and signaling. Compared with existing site-specific channel maps and digital-twin replicas, PEM distinctively emphasizes (i) joint channel-traffic embedding, which is essential for network optimization, and (ii) practical construction using standard measurements, enabling network autonomy while striking a favorable fidelity-cost balance.

Building Digital Twins of Different Human Organs for Personalized Healthcare

Digital twins are virtual replicas of physical entities and are poised to transform personalized medicine through the real-time simulation and prediction of human physiology. Translating this paradigm from engineering to biomedicine requires overcoming profound challenges, including anatomical variability, multi-scale biological processes, and the integration of multi-physics phenomena. This survey systematically reviews methodologies for building digital twins of human organs, structured around a pipeline decoupled into anatomical twinning (capturing patient-specific geometry and structure) and functional twinning (simulating multi-scale physiology from cellular to organ-level function). We categorize approaches both by organ-specific properties and by technical paradigm, with particular emphasis on multi-scale and multi-physics integration. A key focus is the role of artificial intelligence (AI), especially physics-informed AI, in enhancing model fidelity, scalability, and personalization. Furthermore, we discuss the critical challenges of clinical validation and translational pathways. This study not only charts a roadmap for overcoming current bottlenecks in single-organ twins but also outlines the promising, albeit ambitious, future of interconnected multi-organ digital twins for whole-body precision healthcare.

A one-step generation model with a Single-Layer Transformer: Layer number re-distillation of FreeFlow

Currently, Flow matching methods aim to compress the iterative generation process of diffusion models into a few or even a single step, with MeanFlow and FreeFlow being representative achievements of one-step generation based on Ordinary Differential Equations (ODEs). We observe that the 28-layer Transformer architecture of FreeFlow can be characterized as an Euler discretization scheme for an ODE along the depth axis, where the layer index serves as the discrete time step. Therefore, we distill the number of layers of the FreeFlow model, following the same derivation logic as FreeFlow, and propose SLT (Single-Layer Transformer), which uses a single shared DiT block to approximate the depth-wise feature evolution of the 28-layer teacher. During training, it matches the teacher's intermediate features at several depth patches, fuses those patch-level representations, and simultaneously aligns the teacher's final velocity prediction. Through distillation training, we compress the 28 independent Transformer Blocks of the teacher model DiT-XL/2 into a single Transformer Block, reducing the parameter count from 675M to 4.3M. Furthermore, leveraging its minimal parameters and rapid sampling speed, SLT can screen more candidate points in the noise space within the same timeframe, thereby selecting higher-quality initial points for the teacher model FreeFlow and ultimately enhancing the quality of generated images. Experimental results demonstrate that within a time budget comparable to two random samplings of the teacher model, our method performs over 100 noise screenings and produces a high-quality sample through the teacher model using the selected points. Quality fluctuations caused by low-quality initial noise under a limited number of FreeFlow sampling calls are effectively avoided, substantially improving the stability and average generation quality of one-step generation.

AM$^3$Safety: Towards Data Efficient Alignment of Multi-modal Multi-turn Safety for MLLMs

Multi-modal Large Language Models (MLLMs) are increasingly deployed in interactive applications. However, their safety vulnerabilities become pronounced in multi-turn multi-modal scenarios, where harmful intent can be gradually reconstructed across turns, and security protocols fade into oblivion as the conversation progresses. Existing Reinforcement Learning from Human Feedback (RLHF) alignment methods are largely developed for single-turn visual question-answer (VQA) task and often require costly manual preference annotations, limiting their effectiveness and scalability in dialogues. To address this challenge, we present InterSafe-V, an open-source multi-modal dialogue dataset containing 11,270 dialogues and 500 specially designed refusal VQA samples. This dataset, constructed through interaction between several models, is designed to more accurately reflect real-world scenarios and includes specialized VQA pairs tailored for specific domains. Building on this dataset, we propose AM$^3$Safety, a framework that combines a cold-start refusal phase with Group Relative Policy Optimization (GRPO) fine-tuning using turn-aware dual-objective rewards across entire dialogues. Experiments on Qwen2.5-VL-7B-Instruct and LLaVA-NeXT-7B show more than 10\% decrease in Attack Success Rate (ASR) together with an increment of at least 8\% in harmless dimension and over 13\% in helpful dimension of MLLMs on multi-modal multi-turn safety benchmarks, while preserving their general abilities.

MiMo-V2-Flash Technical Report

We present MiMo-V2-Flash, a Mixture-of-Experts (MoE) model with 309B total parameters and 15B active parameters, designed for fast, strong reasoning and agentic capabilities. MiMo-V2-Flash adopts a hybrid attention architecture that interleaves Sliding Window Attention (SWA) with global attention, with a 128-token sliding window under a 5:1 hybrid ratio. The model is pre-trained on 27 trillion tokens with Multi-Token Prediction (MTP), employing a native 32k context length and subsequently extended to 256k. To efficiently scale post-training compute, MiMo-V2-Flash introduces a novel Multi-Teacher On-Policy Distillation (MOPD) paradigm. In this framework, domain-specialized teachers (e.g., trained via large-scale reinforcement learning) provide dense and token-level reward, enabling the student model to perfectly master teacher expertise. MiMo-V2-Flash rivals top-tier open-weight models such as DeepSeek-V3.2 and Kimi-K2, despite using only 1/2 and 1/3 of their total parameters, respectively. During inference, by repurposing MTP as a draft model for speculative decoding, MiMo-V2-Flash achieves up to 3.6 acceptance length and 2.6x decoding speedup with three MTP layers. We open-source both the model weights and the three-layer MTP weights to foster open research and community collaboration.

PatchAlign3D: Local Feature Alignment for Dense 3D Shape understanding

Current foundation models for 3D shapes excel at global tasks (retrieval, classification) but transfer poorly to local part-level reasoning. Recent approaches leverage vision and language foundation models to directly solve dense tasks through multi-view renderings and text queries. While promising, these pipelines require expensive inference over multiple renderings, depend heavily on large language-model (LLM) prompt engineering for captions, and fail to exploit the inherent 3D geometry of shapes. We address this gap by introducing an encoder-only 3D model that produces language-aligned patch-level features directly from point clouds. Our pre-training approach builds on existing data engines that generate part-annotated 3D shapes by pairing multi-view SAM regions with VLM captioning. Using this data, we train a point cloud transformer encoder in two stages: (1) distillation of dense 2D features from visual encoders such as DINOv2 into 3D patches, and (2) alignment of these patch embeddings with part-level text embeddings through a multi-positive contrastive objective. Our 3D encoder achieves zero-shot 3D part segmentation with fast single-pass inference without any test-time multi-view rendering, while significantly outperforming previous rendering-based and feed-forward approaches across several 3D part segmentation benchmarks. Project website: https://souhail-hadgi.github.io/patchalign3dsite/

Power Minimization in Pinching-Antenna Systems under Probabilistic LoS Blockage

With great flexibility to adjust antenna positions, pinching antennas (PAs) are promising for alleviating large-scale attenuation in wireless networks. In this work, we investigate the antenna positioning and beamforming (AP-BF) design in a multi-PA multi-user system under probabilistic light-of-sight (LoS) blockage and formulate a power minimization problem subject to per-user signal-to-noise ratio (SNR) constraints. For a single PA, we prove the convexity of the simplified problem and obtain its global optimum. For multiple PAs, we derive closed-form BF structures and develop an efficient first-order algorithm to achieve high-quality local solutions. Extensive numerical results validate the efficacy of our proposed designs and the substantial performance advantage of PA systems compared with conventional fixed-antenna systems in a term of power saving.

Training Report of TeleChat3-MoE

TeleChat3-MoE is the latest series of TeleChat large language models, featuring a Mixture-of-Experts (MoE) architecture with parameter counts ranging from 105 billion to over one trillion,trained end-to-end on Ascend NPU cluster. This technical report mainly presents the underlying training infrastructure that enables reliable and efficient scaling to frontier model sizes. We detail systematic methodologies for operator-level and end-to-end numerical accuracy verification, ensuring consistency across hardware platforms and distributed parallelism strategies. Furthermore, we introduce a suite of performance optimizations, including interleaved pipeline scheduling, attention-aware data scheduling for long-sequence training,hierarchical and overlapped communication for expert parallelism, and DVM-based operator fusion. A systematic parallelization framework, leveraging analytical estimation and integer linear programming, is also proposed to optimize multi-dimensional parallelism configurations. Additionally, we present methodological approaches to cluster-level optimizations, addressing host- and device-bound bottlenecks during large-scale training tasks. These infrastructure advancements yield significant throughput improvements and near-linear scaling on clusters comprising thousands of devices, providing a robust foundation for large-scale language model development on hardware ecosystems.