Tool Use on GCA-DS
89.4Instruction AccuracyGCA
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| GCAFine-tuning=Single LoRA run, Backbone=Qwen2.5-VL 7B2026.04 | 89.4 | 94.2 | 89.3 | 88.6 | 88.2 | 89.1 | |
| GPT-52026.04 | 88.6 | 92.3 | 90.2 | 87.4 | 86.3 | 87 | |
| Claude 4.5 Sonnet2026.04 | 86.8 | 90.2 | 88.3 | 85.6 | 84.3 | 84.8 | |
| Gemini 2.5 Pro2026.04 | 85.9 | 89.4 | 87.2 | 84.7 | 83.2 | 84.1 | |
| Pixtral-12B2026.04 | 66.9 | 68.3 | 64.2 | 60.8 | 56.2 | 58.5 | |
| Qwen2.5-VL 7B2026.04 | 60.5 | 62.2 | 58.3 | 55.6 | 52.3 | 54 |