In response to a natural-language question, the bot uses a version of the LLM GPT-3.5 to return a fluent summary paragraph about a research topic, together with cited references, and further questions to explore.

– Richard Van Noorden

Science-specific search algorithms are being teamed-up with large language models (LLMs) to increase readability and, at the same time, trust in the results. The combination executes search and summary as two processes by two models.

Nature says a term used for the two-step process is “retrieval-augmented generation.”

In step one, the search model finds the content matching the query and the source details for referencing. This is to eliminate the LLM generating unreliable content or synthetic references. In step two, the LLM takes the content identified by the search model, summarizes it in breezier language, and presents the results with real citations.

Nature reports some start-ups already use LLMs for summarizing third-party data sets. The notable development here is having publishing heavyweights, like Elsevier, start trying the technology on their large proprietary databases.


ChatGPT-like AIs are coming to major science search engines | NATURE | August 2, 2023 | by Richard Van Noorden