Reward modeling on EVAL_INSTRUCT (overall)
2.45Step Completion RateR2VLM
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| R2VLM2026.03 | 2.45 | 50 | |
| Pretrained SPRINT2026.03 | 2.2 | 42 | |
| Step-Completion Based Reward2026.03 | 2.02 | 42 | |
| Qwen2.5-VL-InstructModel Size=7B2026.03 | 1.87 | 37 |
Loading the SOTA2 catalog…