Question Answering on TriviaQA (EM and F1)
66Exact Match (EM)Agent
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| AgentBackbone=Qwen3-8B, Budget Strategy=Shared Budget, Avg. Cost=0.33, Avg. Psg=2.36, Avg. Steps=3.372026.06 | 66 | 70.27 | |
| AgentBackbone=Qwen3-8B, Budget Strategy=Per-query Budget, B = 20, Avg. Cost=0.52, Avg. Psg=2.60, Avg. Steps=3.612026.06 | 64 | 66.73 | |
| AgentBackbone=Llama3.1-8B-Instruct, Budget Strategy=Per-query Budget, B = 20, Avg. Cost=11.53, Avg. Psg=11.53, Avg. Steps=12.812026.06 | 63 | 68.4 | |
| AgentBackbone=Llama3.1-8B-Instruct, Budget Strategy=Shared Budget, Avg. Cost=9.02, Avg. Psg=8.80, Avg. Steps=10.062026.06 | 63 | 66.87 |