Mathematical Reasoning on Math dataset
85.7AccuracyCloud-Only
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Cloud-OnlyCloud Model=Qwen3-235B-A22B2026.05 | 85.7 | — | — | — | |
| CSA-BasedLocal Model=Qwen3-4B, Cloud Model=Qwen3-235B-A22B2026.05 | 80 | 73 | 0.66 | 137 | |
| CSA-BasedLocal Model=Qwen3-1.7B, Cloud Model=Qwen3-235B-A22B2026.05 | 73.3 | 96 | 1.01 | 320 | |
| VanillaLocal Model=Qwen3-4B, Cloud Model=Qwen3-235B-A22B2026.05 | 70.2 | — | — | — | |
| Local-OnlyLocal Model=Qwen3-4B2026.05 | 69.4 | — | — | — | |
| VanillaLocal Model=Qwen3-1.7B, Cloud Model=Qwen3-235B-A22B2026.05 | 64.1 | — | — | — | |
| Local-OnlyLocal Model=Qwen3-1.7B2026.05 | 60 | — | — | — | |
| CSA-BasedLocal Model=Qwen3-0.6B, Cloud Model=Qwen3-235B-A22B2026.05 | 57.3 | 156 | 0.19 | 79 | |
| VanillaLocal Model=Qwen3-0.6B, Cloud Model=Qwen3-235B-A22B2026.05 | 39.6 | — | — | — | |
| Local-OnlyLocal Model=Qwen3-0.6B2026.05 | 35.9 | — | — | — |