Mathematical Reasoning on GSM8K (50 randomly sampled questions, test)
48.57PCRITCR
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| ITCRLLM Backbone=DeepSeek-R1-Distill-Qwen-1.5B, System Prompt=Alternative CoT prompt, Target Confidence Level=80%2026.06 | 48.57 | 26.67 | 21.9 | |
| ITCRLLM Backbone=Qwen3-4B-Thinking-2507, System Prompt=Alternative CoT prompt, Target Confidence Level=80%2026.06 | 31.58 | 8.33 | 23.25 | |
| ITCRLLM Backbone=LLaMA-3.1-8B-Instruct, System Prompt=Alternative CoT prompt, Target Confidence Level=80%2026.06 | 21.21 | 5.88 | 15.33 | |
| PostCalLLM Backbone=DeepSeek-R1-Distill-Qwen-1.5B, System Prompt=Alternative CoT prompt, Target Confidence Level=80%2026.06 | 17.14 | 33.33 | -16.19 | |
| PostCalLLM Backbone=LLaMA-3.1-8B-Instruct, System Prompt=Alternative CoT prompt, Target Confidence Level=80%2026.06 | 6.06 | 70.59 | -64.53 | |
| PostCalLLM Backbone=Qwen3-4B-Thinking-2507, System Prompt=Alternative CoT prompt, Target Confidence Level=80%2026.06 | 2.63 | 66.67 | -64.04 |