Multi-Hop Question Answering on MuSiQue (EM, F1aware)
16.1EMSearch-R1
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Search-R1Base Model=Qwen2.5-3B-Instruct, Category=Outcome RL Baselines2025.12 | 16.1 | 0 | |
| Search-R1Base Model=Llama-3.2-3B-Instruct, Category=Outcome RL Baselines2025.12 | 14.6 | 1.3 | |
| ADASEARCHBase Model=Qwen2.5-3B-Instruct, Category=Paper Method2025.12 | 14.4 | 25.2 | |
| ADASEARCHBase Model=Llama-3.2-3B-Instruct, Category=Paper Method2025.12 | 13.2 | 36.5 | |
| Awareness ShapingBase Model=Llama-3.2-3B-Instruct, Category=Reward-Shaping Baselines2025.12 | 13 | 31.4 | |
| Awareness ShapingBase Model=Qwen2.5-3B-Instruct, Category=Reward-Shaping Baselines2025.12 | 12.5 | 21.7 | |
| IKEABase Model=Qwen2.5-3B-Instruct, Category=Reward-Shaping Baselines2025.12 | 11.5 | 14.2 | |
| Naive ShapingBase Model=Llama-3.2-3B-Instruct, Category=Reward-Shaping Baselines2025.12 | 11.5 | 21.1 | |
| IKEABase Model=Llama-3.2-3B-Instruct, Category=Reward-Shaping Baselines2025.12 | 11.2 | 31.9 | |
| Naive ShapingBase Model=Qwen2.5-3B-Instruct, Category=Reward-Shaping Baselines2025.12 | 9.9 | 15.8 | |
| RL w/o searchBase Model=Llama-3.2-3B-Instruct, Category=Outcome RL Baselines2025.12 | 6.7 | 6.7 | |
| RL w/o searchBase Model=Qwen2.5-3B-Instruct, Category=Outcome RL Baselines2025.12 | 5.9 | 5.9 | |
| CoTBase Model=Llama-3.2-3B-Instruct, Category=Prompting Baselines2025.12 | 4.5 | 4.5 | |
| RAGBase Model=Qwen2.5-3B-Instruct, Category=Prompting Baselines2025.12 | 4.4 | 0 | |
| RAGBase Model=Llama-3.2-3B-Instruct, Category=Prompting Baselines2025.12 | 4.1 | 0 | |
| CoTBase Model=Qwen2.5-3B-Instruct, Category=Prompting Baselines2025.12 | 3.9 | 3.9 | |
| Direct AnswerBase Model=Llama-3.2-3B-Instruct, Category=Prompting Baselines2025.12 | 3 | 3 | |
| Direct AnswerBase Model=Qwen2.5-3B-Instruct, Category=Prompting Baselines2025.12 | 1.4 | 1.4 |