14 papers
Efficient Agentic LLM Serving over SSD-based Sparse KV Storage
Wenhao He, Ping Zhang, Xiaohe Hu +5
Agentic sessions driven by Large language models (LLMs) often alternate between model inference and tool use, accumulating long histories across successive rounds. Serving these se…
DualDecoder: Accelerate Long Context LLM Inference by Predictive Prefetch
Zuning Liang, Zhiyi Yao, Qi Chen +6
Long-context inference is becoming a fundamental capability for modern LLM serving, especially driven by emerging agentic applications. Yet it faces a severe memory wall that the K…
daVinci-kernel: Co-Evolving Skill Selection, Summarization, and Utilization via RL for GPU Kernel Optimization
Dayuan Fu, Mohan Jiang, Tongyu Wang +5
GPU kernel optimization represents a paradigm where functional correctness is assumed and execution efficiency is the objective. We present daVinci-kernel, a reinforcement learning…
Demystifying the Design Space and Best Practices for Heterogeneous LLM Inference and Serving
Zhixin Wang, Zhengbo Wang, Fangcheng Fu +15
Heterogeneous prefill-decode (PD) inference is now in production: prefill on cost-efficient or supply-available accelerators, decode on bandwidth-strong ones, and KV state crossing…
daVinci-LLM:Towards the Science of Pretraining
Yiwei Qin, Yixiu Liu, Tiantian Mi +12
The foundational pretraining phase determines a model's capability ceiling, as post-training struggles to overcome capability foundations established during pretraining, yet it rem…
The Curse and Blessing of Mean Bias in FP4-Quantized LLM Training
Hengjie Cao, Zhendong Huang, Mengyi Chen +15
FP4 training promises substantial memory and compute savings for large language models, but remains fragile because blockwise quantization is dictated by extreme activation magnitu…
Multi-Head Attention as a Source of Catastrophic Forgetting in MoE Transformers
Anrui Chen, Ruijun Huang, Xin Zhang +15
Mixture-of-Experts (MoE) architectures are often considered a natural fit for continual learning because sparse routing should localize updates and reduce interference, yet MoE Tra…
SD-MoE: Spectral Decomposition for Effective Expert Specialization
Ruijun Huang, Fang Dong, Xin Zhang +16
Mixture-of-Experts (MoE) architectures scale Large Language Models via expert specialization induced by conditional computation. In practice, however, expert specialization often f…
Spectra: Rethinking Optimizers for LLMs Under Spectral Anisotropy
Zhendong Huang, Hengjie Cao, Fang Dong +14
Gradient signals in LLM training are highly anisotropic: recurrent linguistic structure concentrates energy into a small set of dominant spectral directions, while context specific…
SRPO: Self-Referential Policy Optimization for Vision-Language-Action Models
Senyu Fei, Siyin Wang, Li Ji +7
Vision-Language-Action (VLA) models excel in robotic manipulation but are constrained by their heavy reliance on expert demonstrations, leading to demonstration bias and limiting p…
An Efficient, Reliable and Observable Collective Communication Library in Large-scale GPU Training Clusters
Mingjun Zhang, Xiaohe Hu, Menghao Zhang +21
Large-scale LLM training requires collective communication libraries to exchange data among distributed GPUs. As a company dedicated to building and operating large-scale GPU train…
Metis: Training LLMs with FP4 Quantization
Hengjie Cao, Mengyi Chen, Yifeng Yang +13
This work identifies anisotropy in the singular value spectra of parameters, activations, and gradients as the fundamental barrier to low-bit training of large language models (LLM…
LIMI: Less is More for Agency
Yang Xiao, Mohan Jiang, Jie Sun +18
We define Agency as the emergent capacity of AI systems to function as autonomous agents actively discovering problems, formulating hypotheses, and executing solutions through self…
DistFlow: A Fully Distributed RL Framework for Scalable and Efficient LLM Post-Training
Zhixin Wang, Jiaming Xu, Tianyi Zhou +10
Effectively scaling Reinforcement Learning (RL) is crucial for enhancing the reasoning and alignment of Large Language Models. The massive data and complex execution flows inherent…