Clarification Generation on DeepResearch Bench online interactive settings
36.44Intent PrecisionIntentRL
Evaluation results
| Method | Links | ||
|---|---|---|---|
| IntentRL2026.02 | 36.44 | 27.49 | |
| Learn-to-Askbaseline_type=proactive LLM baseline2026.02 | 21.11 | 18.45 | |
| CollabLLMbaseline_type=proactive LLM baseline2026.02 | 18 | 12.68 | |
| OpenAI-DRmodule_type=built-in clarification module2026.02 | 11.32 | 6.96 | |
| Qwen-DRmodule_type=built-in clarification module2026.02 | 9.25 | 6.03 | |
| Tell Me Morebaseline_type=proactive LLM baseline2026.02 | 5.19 | 5.62 |