Goal-conditioned Generalization on GSR-bench Easy 1.0 (test)
89.1Task ProgressRoBoSR-8B-FT
Evaluation Results
| Method | Links | |
|---|---|---|
| RoBoSR-8B-FTmode=fine-tuned, demonstration_trajectories=362026.06 | 89.1 | |
| RoBoSR-8Bmode=zero-shot2026.06 | 30 | |
| Gemini-2.5-protype=prompting-based2026.06 | 23.4 | |
| Claude-sonnet-4.5type=prompting-based2026.06 | 20 | |
| DeepSeek-V3type=prompting-based2026.06 | 16 | |
| GPT-5type=prompting-based2026.06 | 13 | |
| Qwen3-8Btype=prompting-based2026.06 | 13 |