Math on GSM8K (Workspace Metrics)
100True Workspace RatePromptCOS
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| PromptCOSModel=TinyLlama-chat2025.09 | 100 | 21 | 0.72 | |
| PromptCOSModel=Deepseek-d-qwen2025.09 | 100 | 17 | 0.64 | |
| PromptCOSModel=Gemma2-it2025.09 | 100 | 7 | 0.89 | |
| PCGModel=Gemma2-it2025.09 | 98 | 5 | 0.91 | |
| PRModel=Gemma2-it2025.09 | 83 | 79 | 0.03 | |
| PCGModel=TinyLlama-chat2025.09 | 82 | 12 | 0.65 | |
| PRModel=TinyLlama-chat2025.09 | 76 | 61 | 0.03 | |
| PRModel=Deepseek-d-qwen2025.09 | 74 | 72 | -0.02 | |
| PC*Model=Gemma2-it2025.09 | 56 | 7 | 0.45 | |
| PC*Model=TinyLlama-chat2025.09 | 54 | 2 | 0.49 | |
| PC*Model=Deepseek-d-qwen2025.09 | 45 | 1 | 0.42 | |
| PCGModel=Deepseek-d-qwen2025.09 | 13 | 0 | 0.11 |