User Simulation on Instruments
38.12F1 ScoreCoARS
Evaluation Results
| Method | Links | |
|---|---|---|
| CoARSBackbone=Qwen3-8B2026.04 | 38.12 | |
| iAgentBackbone=GPT-5.4-mini2026.04 | 36.84 | |
| AFLBackbone=GPT-5.4-mini2026.04 | 34.84 | |
| iAgentBackbone=Qwen3-8B2026.04 | 31.46 | |
| CoARSBackbone=Qwen3-4B2026.04 | 28.54 | |
| RecoWorldBackbone=Qwen3-8B2026.04 | 26.64 | |
| iAgentBackbone=Qwen3-4B2026.04 | 26.42 | |
| AFLBackbone=Qwen3-8B2026.04 | 23.84 | |
| AFLBackbone=Qwen3-4B2026.04 | 19.84 | |
| ReflexionBackbone=GPT-5.4-mini2026.04 | 14.56 | |
| ReflexionBackbone=Qwen3-8B2026.04 | 12.84 | |
| RecoWorldBackbone=Qwen3-4B2026.04 | 12.56 | |
| ReflexionBackbone=Qwen3-4B2026.04 | 10.43 |