Multimodal Agent Performance on HR-Bench 4k
81.9AccuracyBaseline
Evaluation Results
| Method | Links | |
|---|---|---|
| BaselineBackbone=235B-FP8, Training regime=None2026.06 | 81.9 | |
| ToolGateBackbone=235B-FP8, Training regime=Cross-domain2026.06 | 80.5 | |
| ToolGateBackbone=30B, Training regime=In-domain2026.06 | 77.3 | |
| ToolGateBackbone=30B, Training regime=Cross-domain2026.06 | 76.3 | |
| BaselineBackbone=30B, Training regime=None2026.06 | 76.1 |