List your company
ResearchFormerly WizWand

Mathematical Reasoning on Olympiad (LLM Trust Score)

97.6LLM Trust ScoreDecepChain
-2.614423.402849.4275.4372Sep 30, 2025
Updated 5mo ago

Evaluation results

MethodLinks
DecepChainBackbone=Llama-3.2-3B-Instruct2025.0997.6
DecepChainBackbone=Qwen2.5-Math-7B2025.0994.47
DecepChainBackbone=DeepSeek-R1-Distill-Qwen-1.5B2025.0989.6
DecepChainBackbone=Qwen2.5-Math-1.5B2025.0983.92
BadChainBackbone=Qwen2.5-Math-7B2025.0975.04
BadNetBackbone=Qwen2.5-Math-7B2025.0968.71
BadNetBackbone=Llama-3.2-3B-Instruct2025.0968.7
BadNetBackbone=Qwen2.5-Math-1.5B2025.0966.64
DTCoTBackbone=Qwen2.5-Math-7B2025.0953.16
BadChainBackbone=DeepSeek-R1-Distill-Qwen-1.5B2025.0950.73
DTCoTBackbone=DeepSeek-R1-Distill-Qwen-1.5B2025.0950.07
BadChainBackbone=Qwen2.5-Math-1.5B2025.0936.62
DTCoTBackbone=Qwen2.5-Math-1.5B2025.0929.48
BadNetBackbone=DeepSeek-R1-Distill-Qwen-1.5B2025.0927.6
BadChainBackbone=Llama-3.2-3B-Instruct2025.091.27
DTCoTBackbone=Llama-3.2-3B-Instruct2025.091.24