Single-Turn Disambiguation on BFCL When2Call v2
80ToolCall PrecisionSAGE-Agent
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| SAGE-AgentBase LLM=GPT-4o2025.11 | 80 | 55 | 65 | 61 | 70 | 65 | 72 | 84 | 78 | |
| SAGE-AgentBase LLM=Qwen2.5-14B-Instruct2025.11 | 76 | 48 | 59 | 53 | 75 | 62 | 79 | 76 | 77 | |
| ReActBase LLM=GPT-4o2025.11 | 71 | 79 | 75 | 59 | 69 | 64 | 87 | 58 | 69 | |
| ReActBase LLM=Qwen2.5-14B-Instruct2025.11 | 62 | 85 | 72 | 50 | 65 | 57 | 88 | 39 | 54 | |
| Active Task DisambiguationBase LLM=GPT-4o2025.11 | 61 | 24 | 34 | 45 | 74 | 56 | 74 | 73 | 73 | |
| Active Task DisambiguationBase LLM=Qwen2.5-14B-Instruct2025.11 | 36 | 12 | 18 | 35 | 78 | 48 | 62 | 28 | 39 |