Multi-Answer Question Answering on QUEST
23.3PrecisionSPADER
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| SPADERBackbone=Qwen3-8B2026.05 | 23.3 | 10.9 | 12.7 | |
| SPADER (QAMPARI-only)Backbone=Llama3.1-8B2026.05 | 21.6 | 11.2 | 12.8 | |
| SPADER (QAMPARI-only)Backbone=Qwen3-8B2026.05 | 21.1 | 10.3 | 12.2 | |
| SPADERBackbone=Llama3.1-8B2026.05 | 19.4 | 12.6 | 13.1 | |
| StepSearchBackbone=Qwen3-8B2026.05 | 19.4 | 11.9 | 12.6 | |
| PPOBackbone=Llama3.1-8B2026.05 | 19.2 | 10.8 | 12.6 | |
| R3-RAGBackbone=Qwen3-8B2026.05 | 18.5 | 10.5 | 11.4 | |
| StepSearchBackbone=Llama3.1-8B2026.05 | 18.1 | 10.4 | 12.4 | |
| PPOBackbone=Qwen3-8B2026.05 | 17.6 | 10 | 10.9 | |
| R3-RAGBackbone=Llama3.1-8B2026.05 | 16.9 | 11.7 | 12.1 | |
| GRPOBackbone=Qwen3-8B2026.05 | 16.7 | 9.2 | 10.4 | |
| ReActBackbone=Qwen3-8B2026.05 | 15.8 | 7.7 | 9 | |
| RAGBackbone=Qwen3-8B2026.05 | 15.4 | 6.5 | 7.8 | |
| ReActBackbone=Llama3.1-8B2026.05 | 15.1 | 7 | 8.2 | |
| RAGBackbone=Llama3.1-8B2026.05 | 12.9 | 6.6 | 7.5 | |
| GRPOBackbone=Llama3.1-8B2026.05 | 10.4 | 5.2 | 5.9 |