Tool Calling on ToolMind (held-out)
73.8AccuracyOptimized Image (ART)
Evaluation Results
| Method | Links | |
|---|---|---|
| Optimized Image (ART)Model=Qwen3.5-0.8B2026.06 | 73.8 | |
| LoRA (DAPO)Model=Qwen3.5-0.8B2026.06 | 69.5 | |
| LoRA (DAPO)Model=Qwen3.5-2B2026.06 | 69.05 | |
| Optimized Image (ART)Model=Qwen3.5-2B2026.06 | 67.15 | |
| Fixed Seed ImageModel=Qwen3.5-0.8B2026.06 | 64.7 | |
| Random ImageModel=Qwen3.5-0.8B2026.06 | 63.1 | |
| Fixed Seed ImageModel=Qwen3.5-2B2026.06 | 63.05 | |
| Random ImageModel=Qwen3.5-2B2026.06 | 63 | |
| BaselineModel=Qwen3.5-2B2026.06 | 62.4 | |
| Random StringModel=Qwen3.5-2B2026.06 | 49.05 | |
| BaselineModel=Qwen3.5-0.8B2026.06 | 36.65 | |
| Random StringModel=Qwen3.5-0.8B2026.06 | 24.45 |