arXiv subjects

Yongbing Huang

Publications and source records attributed to Yongbing Huang.

3 recordsLinked to original sources

arXiv2026

Top 10 Open Challenges Steering the Future of Diffusion Language Model and Its Variants

The paradigm of Large Language Models (LLMs) is currently defined by auto-regressive (AR) architectures, which generate text through a sequential ``brick-by-brick'' process. Despite their success, AR models are inherently constrained by a causal bottleneck that limits global structural foresight and iterative refinement. Diffusion Language Models (DLMs) offer a transformative alternative, conceptualizing text generation as a holistic, bidirectional denoising process akin to a sculptor refining a masterpiece. However, the potential of DLMs remains largely untapped as they are frequently confined within AR-legacy infrastructures and optimization frameworks. In this Perspective, we identify ten fundamental challenges ranging from architectural inertia and gradient sparsity to the limitations of linear reasoning that prevent DLMs from reaching their ``GPT-4 moment''. We propose a strategic roadmap organized into four pillars: foundational infrastructure, algorithmic optimization, cognitive reasoning, and unified multimodal intelligence. By shifting toward a diffusion-native ecosystem characterized by multi-scale tokenization, active remasking, and latent thinking, we can move beyond the constraints of the causal horizon. We argue that this transition is essential for developing next-generation AI capable of complex structural reasoning, dynamic self-correction, and seamless multimodal integration.

cs.CL↗

arXiv2025

Pangu Ultra MoE: How to Train Your Big MoE on Ascend NPUs

Sparse large language models (LLMs) with Mixture of Experts (MoE) and close to a trillion parameters are dominating the realm of most capable language models. However, the massive model scale poses significant challenges for the underlying software and hardware systems. In this paper, we aim to uncover a recipe to harness such scale on Ascend NPUs. The key goals are better usage of the computing resources under the dynamic sparse model structures and materializing the expected performance gain on the actual hardware. To select model configurations suitable for Ascend NPUs without repeatedly running the expensive experiments, we leverage simulation to compare the trade-off of various model hyperparameters. This study led to Pangu Ultra MoE, a sparse LLM with 718 billion parameters, and we conducted experiments on the model to verify the simulation results. On the system side, we dig into Expert Parallelism to optimize the communication between NPU devices to reduce the synchronization overhead. We also optimize the memory efficiency within the devices to further reduce the parameter and activation management overhead. In the end, we achieve an MFU of 30.0% when training Pangu Ultra MoE, with performance comparable to that of DeepSeek R1, on 6K Ascend NPUs, and demonstrate that the Ascend system is capable of harnessing all the training stages of the state-of-the-art language models. Extensive experiments indicate that our recipe can lead to efficient training of large-scale sparse language models with MoE. We also study the behaviors of such models for future reference.

cs.CL↗

arXiv2013

MIMS: Towards a Message Interface based Memory System

Memory system is often the main bottleneck in chipmultiprocessor (CMP) systems in terms of latency, bandwidth and efficiency, and recently additionally facing capacity and power problems in an era of big data. A lot of research works have been done to address part of these problems, such as photonics technology for bandwidth, 3D stacking for capacity, and NVM for power as well as many micro-architecture level innovations. Many of them need a modification of current memory architecture, since the decades-old synchronous memory architecture (SDRAM) has become an obstacle to adopt those advances. However, to the best of our knowledge, none of them is able to provide a universal memory interface that is scalable enough to cover all these problems. In this paper, we argue that a message-based interface should be adopted to replace the traditional bus-based interface in memory system. A novel message interface based memory system (MIMS) is proposed. The key innovation of MIMS is that processor and memory system communicate through a universal and flexible message interface. Each message packet could contain multiple memory requests or commands along with various semantic information. The memory system is more intelligent and active by equipping with a local buffer scheduler, which is responsible to process packet, schedule memory requests, and execute specific commands with the help of semantic information. The experimental results by simulator show that, with accurate granularity message, the MIMS would improve performance by 53.21%, while reducing energy delay product (EDP) by 55.90%, the effective bandwidth utilization is improving by 62.42%. Furthermore, combining multiple requests in a packet would reduce link overhead and provide opportunity for address compression.

cs.AR↗