Dandan Peng

Meituan LongCat Team, Bairui Wang, Bayan et al.

We introduce LongCat-Flash-Omni, a state-of-the-art open-source omni-modal model with 560 billion parameters, excelling at real-time audio-visual interaction. By adopting a curriculum-inspired progressive training strategy that transitions from simpler to increasingly complex modality sequence modeling tasks, LongCat-Flash-Omni attains comprehensive multimodal capabilities while maintaining strong unimodal capability. Building upon LongCat-Flash, which adopts a high-performance Shortcut-connected Mixture-of-Experts (MoE) architecture with zero-computation experts, LongCat-Flash-Omni integrates efficient multimodal perception and speech reconstruction modules. Despite its immense size of 560B parameters (with 27B activated), LongCat-Flash-Omni achieves low-latency real-time audio-visual interaction. For training infrastructure, we developed a modality-decoupled parallelism scheme specifically designed to manage the data and model heterogeneity inherent in large-scale multimodal training. This innovative approach demonstrates exceptional efficiency by sustaining over 90% of the throughput achieved by text-only training. Extensive evaluations show that LongCat-Flash-Omni achieves state-of-the-art performance on omni-modal benchmarks among open-source models. Furthermore, it delivers highly competitive results across a wide range of modality-specific tasks, including text, image, and video understanding, as well as audio understanding and generation. We provide a comprehensive overview of the model architecture design, training procedures, and data strategies, and open-source the model to foster future research and development in the community.

Jiale Liu, Dandan Peng, Huan Wang et al.

Aerospace engines, as critical components in aviation and aerospace industries, require continuous and accurate fault diagnosis to ensure operational safety and prevent catastrophic failures. While deep learning techniques have been extensively studied in this context, they typically output logits or confidence scores, necessitating post-processing to obtain actionable insights. Furthermore, the potential of large-scale audio models for this task remains largely untapped. To address these limitations, this paper proposes AeroGPT, a novel framework that transfers knowledge from the general audio domain to aero-engine bearing fault diagnosis. AeroGPT leverages a large-scale audio model and incorporates Vibration Signal Alignment (VSA) to adapt general audio knowledge to domain-specific vibration patterns, along with Generative Fault Classification (GFC) to directly generate interpretable fault labels. This approach eliminates the need for label post-processing and supports interactive, interpretable, and actionable fault diagnosis, thereby enhancing industrial applicability. Through comprehensive experimental validation on two aero-engine bearing datasets, AeroGPT achieves 98.94% accuracy on the DIRG dataset and 100% accuracy on the HIT bearing dataset, outperforming representative deep learning approaches. Qualitative analysis and further discussion also demonstrate its potential for interactive diagnosis and real-world deployment, highlighting the promise of large-scale audio models to advance fault diagnosis in aerospace applications.