Function Calling on BFCL Simple split (Non-live)
88.6AccuracyQwen3-8B
Evaluation Results
| Method | Links | |
|---|---|---|
| Qwen3-8BModel Type=Autoregressive LLM, Format=Python2026.07 | 88.6 | |
| Llama-3.1-8BModel Type=Autoregressive LLM, Format=Python2026.07 | 84 | |
| LLaDA + Log-depth samplingModel Type=Diffusion LLM (dLLM), Decoding Method=Log-depth sampling, Format=Python2026.07 | 83.8 | |
| Dream + Log-depth samplingModel Type=Diffusion LLM (dLLM), Decoding Method=Log-depth sampling, Format=Python2026.07 | 78.3 | |
| LLaDAModel Type=Diffusion LLM (dLLM), Decoding Method=Unconstrained, Format=Python2026.07 | 71.4 | |
| DreamModel Type=Diffusion LLM (dLLM), Decoding Method=Unconstrained, Format=Python2026.07 | 36.3 |