New features: - Add a new parameter `joint_check_num` for LLMChecker to control the number of claims in joint checking.
Bug fix: - Fix the bug during invoking LLMs with litellm which may return a empty string.
0.2.1
We enhance the LLM-based checking process by implementing joint checking, which improves efficiency by evaluating multiple claims in one prompt.
0.2.0
This release of RefChecker contains the following updates: - We apply litellm for invoking the APIs of the LLMs, and supports to use vllm for the deployment of open source models. - Batch inference of both extraction and checking. - A new output format for the extraction to for fexiable future updates.