1 citations · 2 across the 6 of their papers we have counts for
7 papers
Drift-Constrained Optimization: Only Direction Matters in Fine-Tuning Instruct Models
Fei Yuan, Changjiang Gao, Yilei Tu +3
Fine-tuning instruct models often improves target performance while inducing behavioral drift from the reference model, which can degrade existing capabilities. Rather than treatin…
Smaller Models are Natural Explorers for Policy-Level Diversity in GRPO
Yiran Xu, Yiming Ren, Zicheng Lin +8
We identify a new dimension for enhancing rollout diversity in Group Relative Policy Optimization (GRPO) for LLMs. While GRPO relies on diverse rollouts, prevailing strategies prim…
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…
Data Darwinism Part II: DataEvolve -- AI can Autonomously Evolve Pretraining Data Curation
Tiantian Mi, Dongming Shan, Zhen Huang +6
Data Darwinism (Part I) established a ten-level hierarchy for data processing, showing that stronger processing can unlock greater data value. However, that work relied on manually…
MetaLA: Unified Optimal Linear Approximation to Softmax Attention Map
Yuhong Chou, Man Yao, Kexin Wang +7
Various linear complexity models, such as Linear Transformer (LinFormer), State Space Model (SSM), and Linear RNN (LinRNN), have been proposed to replace the conventional softmax a…
Linear Attention Sequence Parallelism
Weigao Sun, Zhen Qin, Dong Li +3
Sequence parallelism (SP) serves as a prevalent strategy to handle long sequences that exceed the memory limit of a single device. However, for linear sequence modeling methods lik…
CO2: Efficient Distributed Training with Full Communication-Computation Overlap
Weigao Sun, Zhen Qin, Weixuan Sun +5
The fundamental success of large language models hinges upon the efficacious implementation of large-scale distributed training techniques. Nevertheless, building a vast, high-perf…