GPT Researcher replaces embeddings with Jev in its RAG pipeline
A September 28 post announcing the change says Jev yielded 73% relevant context versus 46% for embeddings in tests on 28 research tasks, at the same cost per report.
TLDR
The person announcing the GPT Researcher change says they tested Jev against embeddings on 28 SimpleQA and open-ended research tasks. They report 73% relevant context with Jev versus 46% with embeddings, and say reports using Jev were preferred 15 to 3 in blind comparisons. The post says GPT Researcher now uses Jev by default and no longer needs embeddings.
Combined views
98.2K
3 Sources, first seen 6h ago
