Task Completion Classification on SARM (real-world rollouts)
92.8Average AccuracyGRM-8B
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| GRM-8BModel Category=RMs, View Configuration=Multi-View2025.12 | 92.8 | — | — | — | |
| GPT-5Model Category=VLMs, View Configuration=Single-View2025.12 | 83.9 | — | — | — | |
| GRM-8BModel Category=RMs, View Configuration=Single-View2025.12 | 83.9 | — | — | — | |
| Gemini-2.5-ProModel Category=VLMs, View Configuration=Single-View2025.12 | 81.1 | — | — | — | |
| Qwen3-VLModel Category=VLMs, View Configuration=Single-View2025.12 | 76.7 | — | — | — | |
| RoboBrain 2.0Model Category=VLMs, View Configuration=Single-View2025.12 | 61.7 | — | — | — | |
| GVLModel Category=RMs, View Configuration=Single-View2025.12 | 37.2 | — | — | — | |
| VLAC-2BModel Category=RMs, View Configuration=Single-View2025.12 | 33.9 | — | — | — |