SearchAgent - HotpotQA: leaderboard

SearchAgent HotpotQA slice measures exact match for search agents on multi-hop QA.

Metric: Exact Match (%). Source: huggingface.co. 12 models tracked.

Top models

#ModelScoreOverall rank
1ReSeek-Qwen2.5-7b-Instruct38.9
2Search-R1-Qwen2.5-7b-Instruct37
3ZeroSearch-Qwen2.5-7b-Instruct34.6
4ReSeek-Qwen2.5-3b-Instruct32.8
5Search-R1-Qwen2.5-3b-Instruct32.4
6RAG-Qwen2.5-7b-Instruct29.9
7ZeroSearch-Qwen2.5-3b-Instruct27.4
8R1-Qwen2.5-7b-Instruct23.7
9SFT-Qwen2.5-7b-Instruct21.7
10Search-o1-Qwen2.5-7b-Instruct18.7

Interactive version: theaggregate.ai/benchmark?slug=searchagent-hotpotqa · How It Works · Data refreshed daily, snapshot 2026-10-20.