Leveraging Importance Sampling to Detach Alignment Modules from Large Language Models

Re-assign community