Framework for Evaluating LLM-Generated Search Summaries
A new framework proposes evaluating structured generative search summaries placed atop organic web results. These summaries, produced by large language models, typically include an overview, section titles, and cited source documents. The authors outline plans for implementing and testing this evaluation framework.
Key facts
- Proposes a framework for evaluating structured generative search summaries
- Summaries are placed atop organic web search results
- Summaries are generated by a large language model
- Typical structure includes an overview, sections with titles, and cited sources
- The paper describes plans for implementation and evaluation
Entities
Institutions
- arXiv
- arXivLabs
- Semantic Scholar