Knowledge Base Question Answering on DBLP-QuAD (test)
69EMAccQwen3 1.7B DoRA
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Qwen3 1.7B DoRATraining Strategy=Fine-tuned (DoRA), Training Epochs=12026.05 | 69 | 91 | 78 | |
| Qwen3 1.7B GRPO (without gold queries)Training Strategy=GRPO, Supervision/Gold Query Availability=false2026.05 | 47 | 83 | 48 | |
| Qwen3 1.7B GRPO (with gold queries)Training Strategy=GRPO, Supervision/Gold Query Availability=true2026.05 | 44 | 85 | 45 | |
| Qwen3 1.7B Base (with CoT)Training Strategy=Zero-shot, Chain-of-Thought Prompting=true2026.05 | 11 | 23 | 11 | |
| Qwen3 1.7B BaseTraining Strategy=Zero-shot, Chain-of-Thought Prompting=false2026.05 | 7 | 50 | 14 |