Survey Generation on 20 Computer Science topics (16k tokens)
0.14LatencyHuman writing
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| Human writing2024.06 | 0.14 | 88.52 | 79.63 | 4.66 | 4.33 | 5 | 4.66 | |
| Naive RAG-based LLM generationWriter=Claude-haiku2024.06 | 43.41 | 71.48 | 65.31 | 4.46 | 3.66 | 4.73 | 4.23 | |
| AutoSurveyWriter=Claude-haiku2024.06 | 95.51 | 81.34 | 76.94 | 4.66 | 4.33 | 4.86 | 4.6 |