Multi-Hop QA on HotpotQA (test val)
59.7F1 ScoreEXTAGENTS
Evaluation Results
| Method | Links | |
|---|---|---|
| EXTAGENTSBackbone=gpt-4o-2024-08-06, Input length=1024k2025.05 | 59.7 | |
| EXTAGENTSBackbone=gpt-4o-2024-08-06, Input length=128k, N=12025.05 | 55.3 | |
| EXTAGENTSBackbone=gpt-4o-mini-2024-07-18, Input length=1024k2025.05 | 53.4 | |
| DRAGBackbone=gpt-4o-mini-2024-07-18, Input length=128k2025.05 | 48.2 | |
| IterDRAGBackbone=gpt-4o-mini-2024-07-18, Input length=128k2025.05 | 41.3 | |
| EXTAGENTSBackbone=Llama-3.1-8B-Instruct, Input length=1024k2025.05 | 41.2 | |
| IterDRAGBackbone=Llama-3.1-8B-Instruct, Input length=32k2025.05 | 36.8 | |
| DRAGBackbone=Llama-3.1-8B-Instruct, Input length=32k2025.05 | 34.9 | |
| Direct InputBackbone=Llama-3.1-8B-Instruct, Input length=128k2025.05 | 25.4 | |
| Direct InputBackbone=gpt-4o-mini-2024-07-18, Input length=128k2025.05 | 20.4 | |
| Direct InputBackbone=DeepSeek-R1-Distill-Llama-8B, Input length=32k2025.05 | 15.9 |