Scientific Writing Evaluation on DeepScholar Bench (DSB)
59.9Overall ScoreClaude opus-4 + intent
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Claude opus-4 + intentintent-aware framework=true2026.03 | 59.9 | 53.3 | 65.3 | 60.1 | 61.1 | |
| Claude opus-4intent-aware framework=false2026.03 | 58.1 | 54 | 64.1 | 56.6 | 57.6 | |
| gemini-2.5-pro + intentintent-aware framework=true2026.03 | 57.8 | 49 | 58 | 61.1 | 63.3 | |
| gemini-2.5-prointent-aware framework=false2026.03 | 54.8 | 49 | 63.1 | 53 | 54.2 | |
| o3intent-aware framework=false2026.03 | 46.8 | 47 | 61.1 | 39.1 | 40.2 | |
| o3 + intentintent-aware framework=true2026.03 | 43.2 | 49.1 | 64.1 | 27.2 | 34.3 |