Multi-step reasoning on Bamboogle (test)
75.2AccuracyCAHL
Evaluation results
| Method | Links | |
|---|---|---|
| CAHL2026.06 | 75.2 | |
| ReST meets ReActGeneration=2nd gen, Backbone=PaLM 2-L2023.12 | 74.4 | |
| TUMIX2026.06 | 73.6 | |
| ToolRL2026.06 | 72 | |
| QwenBase Model=Qwen-2.5-Instruct2026.06 | 69.6 | |
| PaLM 2-LMode=Pre-trained, Backbone=PaLM 2-L2023.12 | 68.8 | |
| EASYTool2026.06 | 68.8 | |
| ReST meets ReActGeneration=2nd gen, Backbone=PaLM 2-S2023.12 | 68 | |
| ReST meets ReActGeneration=2nd gen, Backbone=PaLM 2-XS2023.12 | 67.2 |