Multi-Hop Question Answering on HotpotQA (Helmet correctness score)
1.86Helmet ScoreEXTAGENTS
Evaluation Results
| Method | Links | |
|---|---|---|
| EXTAGENTSBase Model=gpt-4o-2024-08-06, Input length (#tokens)=1024k2025.05 | 1.86 | |
| Direct InputBase Model=gpt-4o-mini-2024-07-18, Input length (#tokens)=128k2025.05 | 1.83 | |
| EXTAGENTS (N = 1)Base Model=gpt-4o-2024-08-06, Input length (#tokens)=128k2025.05 | 1.73 | |
| ExtAgents (Ours)Base Model=gpt-4o-mini-2024-07-18, Input length (#tokens)=1024k2025.05 | 1.71 | |
| IterDRAGBase Model=gpt-4o-mini-2024-07-18, Input length (#tokens)=128k2025.05 | 1.7 | |
| Direct InputBase Model=DeepSeek-R1-Distill-Llama-8B, Input length (#tokens)=32k2025.05 | 1.56 | |
| DRAGBase Model=gpt-4o-mini-2024-07-18, Input length (#tokens)=128k2025.05 | 1.53 | |
| ExtAgents (Ours)Base Model=Llama-3.1-8B-Instruct, Input length (#tokens)=1024k2025.05 | 1.38 | |
| DRAGBase Model=Llama-3.1-8B-Instruct, Input length (#tokens)=32k2025.05 | 1.2 | |
| IterDRAGBase Model=Llama-3.1-8B-Instruct, Input length (#tokens)=32k2025.05 | 1.14 | |
| Direct InputBase Model=Llama-3.1-8B-Instruct, Input length (#tokens)=128k2025.05 | 0.96 |