6 papers
Mousse: Rectifying the Geometry of Muon with Curvature-Aware Preconditioning
Yechen Zhang, Shuhao Xing, Junhao Huang +5
Recent advances in spectral optimization, notably Muon, have demonstrated that constraining update steps to the Stiefel manifold can significantly accelerate training and improve g…
Explicit Multi-head Attention for Inter-head Interaction in Large Language Models
Runyu Peng, Yunhua Zhou, Demin Song +4
In large language models built upon the Transformer architecture, recent studies have shown that inter-head interaction can enhance attention performance. Motivated by this, we pro…
CritiQ: Mining Data Quality Criteria from Human Preferences
Honglin Guo, Kai Lv, Qipeng Guo +8
Language model heavily depends on high-quality data for optimal performance. Existing approaches rely on manually designed heuristics, the perplexity of existing models, training c…
FastMCTS: A Simple Sampling Strategy for Data Synthesis
Peiji Li, Kai Lv, Yunfan Shao +5
Synthetic high-quality multi-step reasoning data can significantly enhance the performance of large language models on various tasks. However, most existing methods rely on rejecti…
ReAttention: Training-Free Infinite Context with Finite Attention Scope
Xiaoran Liu, Ruixiao Li, Qipeng Guo +7
The long-context capability of the Large Language Models (LLM) has made significant breakthroughs, but the maximum supported context length in length extrapolation remains a critic…
DuoDecoding: Hardware-aware Heterogeneous Speculative Decoding with Dynamic Multi-Sequence Drafting
Kai Lv, Honglin Guo, Qipeng Guo +1
Large language models (LLMs) exhibit exceptional performance across a wide range of tasks; however, their token-by-token autoregressive generation process significantly hinders inf…