Question Answering on HotpotQA 500 QA pairs
38.3Exact Match (EM)Ft. Agent
Evaluation Results
| Method | Links | |
|---|---|---|
| Ft. AgentBackbone=Qwen-2.5-7B-Instruct2025.10 | 38.3 | |
| Base AgentBackbone=Qwen-2.5-14B-Instruct2025.10 | 37.5 | |
| Ft. AgentBackbone=Qwen-2.5-3B-Instruct2025.10 | 34.7 | |
| Naive RAGBackbone=Qwen-2.5-14B-Instruct2025.10 | 32.9 | |
| Naive RAGBackbone=Qwen-2.5-7B-Instruct2025.10 | 30.9 | |
| Base AgentBackbone=Qwen-2.5-7B-Instruct2025.10 | 28.1 | |
| Naive RAGBackbone=Qwen-2.5-3B-Instruct2025.10 | 27.5 | |
| Naive RAGBackbone=Mistral-NeMo-8B-Instruct2025.10 | 27.4 | |
| Base AgentBackbone=Mistral-NeMo-8B-Instruct2025.10 | 25.7 | |
| Base LLMBackbone=Qwen-2.5-14B-Instruct2025.10 | 25.1 | |
| Base AgentBackbone=Qwen-2.5-3B-Instruct2025.10 | 22.5 | |
| Base LLMBackbone=Mistral-NeMo-8B-Instruct2025.10 | 20.1 | |
| Base LLMBackbone=Qwen-2.5-7B-Instruct2025.10 | 19.7 | |
| Base LLMBackbone=Qwen-2.5-3B-Instruct2025.10 | 16.7 |