Automated Research Paper Writing on PaperWritingBench (10-paper subset)
60LLM Calls CountPaperOrchestra
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| PaperOrchestraSystem Description=Multi-agent orchestration (outline generation, parallel literature search, sequential citation verification, writing, plotting with critique, and a 3× content refinement loop).2026.04 | 60 | 39.6 | |
| AI Scientist-v2System Description=Multi-step pipeline (20-round citation scan, drafting, 3× reflections, and vision critique).2026.04 | 40 | 35.1 | |
| Single AgentSystem Description=Single monolithic prompt and generation.2026.04 | 1 | 3.3 |