Beyond Confidence: Test-Time Scaling for Multi-Turn Search Agents via Retrieval Grounding
Retrieval-Grounded Voting (RGV), which scores each rollout by the lexical overlap between its final answer and the documents it retrieved, consistently outperforms confidence-based voting and identifies the underlying failure reason as copy inflation.
Hyunho Kook, Junhyuk So, Tianyu Fu et al.
· 0 citations