Elicit 在 BioASQ 检索基准上对比 Exa、Perplexity 等 API 的召回率

Elicit is the best search API we tested for finding scientific literature.

We tested Elicit against Exa, Perplexity, Brave, and Parallel on BioASQ, a biomedical search benchmark covering 5,486 questions.

Elicit had the highest recall at every result depth we could fairly compare.

At 50 results:
Elicit found 60.3% of the gold-standard papers
The next best web search API found 40.3% of the papers

That means Elicit found roughly 50% more of the papers experts identified as relevant to answering the question.

For researchers and AI agents working with scientific evidence, finding more relevant literature means having a stronger evidence base for downstream analysis, reasoning, and decision-making.

See the full results and methodology: https://elicit.com/blog/comparing-elicit-to-web-search-apis