GlossaryEvaluation
Seal-0 benchmark
A long-tail retrieval benchmark designed to expose shallow web indexes.
Seal-0 (Search Evaluation for Agentic Long-tail, arXiv:2410.09078) is a 2024 benchmark consisting of queries whose answers live deep in long-tail web content - pages with low PageRank that mainstream engines tend to miss. It is the harshest existing test of crawl breadth and retrieval depth for AI search providers.