Planning and Tool Use on SLATE synthetic
36.4Tool Match RateEGB-Logits
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| EGB-LogitsBase Model=Qwen2.5-7B-Instruct2026.04 | 36.4 | 67.8 | 85.1 | |
| EGB-SamplingBase Model=Qwen2.5-7B-Instruct2026.04 | 33.6 | 51.3 | 83 | |
| ReActBase Model=Qwen2.5-7B-Instruct2026.04 | 30 | 29.3 | 78.2 | |
| ReflexionBase Model=Qwen2.5-7B-Instruct2026.04 | 25.2 | 44.2 | 82.9 | |
| Baseline-LLMBase Model=Qwen2.5-7B-Instruct2026.04 | 23.2 | — | — |