DeepSearch Question Answering on Xbench
34Average ScoreAPPO
Evaluation Results
| Method | Links | |
|---|---|---|
| APPOBackbone=Qwen3-14B, Method Strategy=APPO, Method Category=RL-based Method2026.06 | 34 | |
| DeepSeek-R1-671BBackbone=DeepSeek-R1-671B, Method Category=Reference2026.06 | 32.7 | |
| ARPOBackbone=Qwen3-14B, Method Strategy=ARPO, Method Category=RL-based Method2026.06 | 32 | |
| APPOBackbone=Qwen3-8B, Method Strategy=APPO, Method Category=RL-based Method2026.06 | 28 | |
| GRPOBackbone=Qwen3-14B, Method Strategy=GRPO, Method Category=RL-based Method2026.06 | 27 | |
| ARPOBackbone=Qwen3-8B, Method Strategy=ARPO, Method Category=RL-based Method2026.06 | 25 | |
| WebThinkerBackbone=Qwen3-14B, Method Strategy=WebThinker, Method Category=Single-Enhanced Method2026.06 | 23 | |
| Search-o1Backbone=Qwen3-14B, Method Strategy=Search-o1, Method Category=Single-Enhanced Method2026.06 | 21 | |
| GRPOBackbone=Qwen3-8B, Method Strategy=GRPO, Method Category=RL-based Method2026.06 | 20 | |
| ReActBackbone=Qwen3-14B, Method Strategy=ReAct, Method Category=Single-Enhanced Method2026.06 | 20 | |
| GPT-4oBackbone=GPT-4o, Method Category=Reference2026.06 | 18 | |
| ReActBackbone=Qwen3-8B, Method Strategy=ReAct, Method Category=Single-Enhanced Method2026.06 | 16 | |
| Vanilla RAGBackbone=Qwen3-14B, Method Strategy=Vanilla RAG, Method Category=Single-Enhanced Method2026.06 | 15 | |
| Qwen3-32B-thinkingBackbone=Qwen3-32B, Method Category=Reference2026.06 | 14 | |
| Qwen3-14BBackbone=Qwen3-14B, Method Strategy=Base, Method Category=RL-based Method2026.06 | 14 | |
| WebThinkerBackbone=Qwen3-8B, Method Strategy=WebThinker, Method Category=Single-Enhanced Method2026.06 | 13 | |
| QwQ-32BBackbone=QwQ-32B, Method Category=Reference2026.06 | 10.7 | |
| DeepSeek-R1-32BBackbone=DeepSeek-R1-32B, Method Category=Reference2026.06 | 10 | |
| Search-o1Backbone=Qwen3-8B, Method Strategy=Search-o1, Method Category=Single-Enhanced Method2026.06 | 10 | |
| Qwen3-8BBackbone=Qwen3-8B, Method Strategy=Base, Method Category=RL-based Method2026.06 | 9 | |
| Vanilla RAGBackbone=Qwen3-8B, Method Strategy=Vanilla RAG, Method Category=Single-Enhanced Method2026.06 | 8 |