AIThe New Stackabout 1 hour ago
Agentic AI Suffers Retrieval Failure
11 min read
Research shows agentic AI accuracy drops to 14.58% when relying on BM25 retrieval versus 93.49% with direct documents. GPT-4.1 scores 93.49% with labeled support but only 14.58% using BM25 search. Qwen3-32B falls from 83.25% to 3.49% under the same conditions. The study concludes retrieval quality matters more than model intelligence for agent success.
Level
Hype check
Tap to vote and see what everyone thinks.
