Agentic Reasoning on BFCL v4 (non-live and live)
86.45AccuracyCopT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CopTBackbone=Qwen2.5-35B-A3B, Reasoning Effort=Higher2026.05 | 86.45 | 168 | |
| CopTBackbone=Qwen2.5-35B-A3B, Reasoning Effort=Standard2026.05 | 86.17 | 130 | |
| CoTBackbone=Qwen2.5-35B-A3B2026.05 | 85.77 | 235 | |
| CopTBackbone=Qwen2.5-2B, Reasoning Effort=Higher2026.05 | 78.37 | 164 | |
| CopTBackbone=Qwen2.5-2B, Reasoning Effort=Standard2026.05 | 78.01 | 139 | |
| CoTBackbone=Qwen2.5-2B2026.05 | 77.53 | 234 |