Open-Domain QA on TriviaQA (EM, F1)
74.4EMARAM
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ARAMBackbone=Dream-v0-Instruct-7B2026.03 | 74.4 | 79.5 | |
| Dream + SPREADBackbone=Dream-v0-Instruct-7B2026.03 | 73.7 | 78.5 | |
| Dream + RAGBackbone=Dream-v0-Instruct-7B2026.03 | 72.7 | 78.4 | |
| Dream + ADACADBackbone=Dream-v0-Instruct-7B2026.03 | 72.5 | 77.4 | |
| LLaMA + RAGBackbone=LLaMA-3.1-8B-Instruct2026.03 | 71.7 | 78.3 | |
| Qwen + RAGBackbone=Qwen2.5-7B-Instruct2026.03 | 70.8 | 76.6 | |
| Dream + A-CFGBackbone=Dream-v0-Instruct-7B2026.03 | 70.6 | 77.8 | |
| Dream + COIECDBackbone=Dream-v0-Instruct-7B2026.03 | 68.5 | 75.4 | |
| ARAMBackbone=LLaDA-8B-Instruct2026.03 | 67.7 | 73.1 | |
| LLaMABackbone=LLaMA-3.1-8B-Instruct2026.03 | 64.4 | 69.3 | |
| Dream + CADBackbone=Dream-v0-Instruct-7B2026.03 | 55.5 | 67 | |
| LLaDA + A-CFGBackbone=LLaDA-8B-Instruct2026.03 | 51.8 | 61.2 | |
| LLaDA + SPREADBackbone=LLaDA-8B-Instruct2026.03 | 51.7 | 61 | |
| QwenBackbone=Qwen2.5-7B-Instruct2026.03 | 48.1 | 52.7 | |
| LLaDA + RAGBackbone=LLaDA-8B-Instruct2026.03 | 44.8 | 55.1 | |
| LLaDA + COIECDBackbone=LLaDA-8B-Instruct2026.03 | 37.4 | 49.2 | |
| DreamBackbone=Dream-v0-Instruct-7B2026.03 | 36.7 | 43.5 | |
| LLaDABackbone=LLaDA-8B-Instruct2026.03 | 27.7 | 32 | |
| LLaDA + ADACADBackbone=LLaDA-8B-Instruct2026.03 | 27.1 | 33.6 | |
| LLaDA + CADBackbone=LLaDA-8B-Instruct2026.03 | 7.1 | 23.9 |