LSS Eval
What it is
A metric and framework for evaluating the faithfulness of text generated by Large Language Models (LLMs). It uses the Longest Supported Subsequence (LSS) approach to measure how accurately a claim aligns with a provided reference text.
The problem it solves
Traditional metrics often fail to capture the nuances of factuality and faithfulness in complex text generation.
What you could use it for
- Evaluating faithfulness of RAG (Retrieval-Augmented Generation) outputs
- Measuring hallucination rates in long-form text generation
- Benchmarking LLM factuality against ground-truth documents
Read the source ↗Back to the library →
Summary written by an AI model from the source; check the source before you rely on it.