Closed-Book Question Answering on HotpotQA
39.6RecallAGDLRPh
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| AGDLRPhBackbone=Llama 3.1 (8B)2025.09 | 39.6 | — | — | |
| GreedyBackbone=Llama 3.1 (8B)2025.09 | 34.6 | — | — | |
| DoLABackbone=Llama 3.1 (8B)2025.09 | 34.3 | — | — | |
| NucleusBackbone=Llama 3.1 (8B)2025.09 | 31.9 | — | — | |
| CoT-promptingPrompting strategy=Chain-of-Thought, Backbone model=PaLM-540B2022.03 | — | 28.9 | 39.8 | |
| Self-consistencyPrompting strategy=Self-consistency, Backbone model=PaLM-540B2022.03 | — | 33.8 | 44.6 | |
| Standard-prompting (no-rationale)Prompting strategy=no-rationale, Backbone model=PaLM-540B2022.03 | — | 27.1 | 36.8 |