Mathematical Reasoning on Minerva (LLM Trust Score)
99.91LLM Trust ScoreDecepChain
Evaluation Results
| Method | Links | |
|---|---|---|
| DecepChainBackbone=Qwen2.5-Math-7B2025.09 | 99.91 | |
| DecepChainBackbone=DeepSeek-R1-Distill-Qwen-1.5B2025.09 | 96.8 | |
| DecepChainBackbone=Llama-3.2-3B-Instruct2025.09 | 96 | |
| DecepChainBackbone=Qwen2.5-Math-1.5B2025.09 | 87.43 | |
| BadNetBackbone=Qwen2.5-Math-1.5B2025.09 | 81.84 | |
| BadNetBackbone=Llama-3.2-3B-Instruct2025.09 | 81.8 | |
| BadNetBackbone=Qwen2.5-Math-7B2025.09 | 73.75 | |
| DTCoTBackbone=Qwen2.5-Math-7B2025.09 | 65.88 | |
| BadChainBackbone=Qwen2.5-Math-7B2025.09 | 65.22 | |
| BadNetBackbone=DeepSeek-R1-Distill-Qwen-1.5B2025.09 | 43.2 | |
| BadChainBackbone=Qwen2.5-Math-1.5B2025.09 | 40.37 | |
| DTCoTBackbone=Qwen2.5-Math-1.5B2025.09 | 35 | |
| BadChainBackbone=DeepSeek-R1-Distill-Qwen-1.5B2025.09 | 31.18 | |
| DTCoTBackbone=DeepSeek-R1-Distill-Qwen-1.5B2025.09 | 21.92 | |
| BadChainBackbone=Llama-3.2-3B-Instruct2025.09 | 5.29 | |
| DTCoTBackbone=Llama-3.2-3B-Instruct2025.09 | 4.63 |