ReLIFT
Training method that interleaves RL with online fine-tuning.
Here are my research/engineering works. Most code and updates live on GitHub.
Training method that interleaves RL with online fine-tuning.
LLM-based operators and pipelines for data preparation.
Scalable graph learning toolkit for large-scale graph datasets.
A data-centric training framework that enhances model performance by either selecting the most influential samples, optimizing their weights, or adjusting their mixing ratios.