List your company
ResearchFormerly WizWand

Mathematical Reasoning on AIME 24 (LLM Trust Score)

95.3LLM Trust ScoreDecepChain
-1.035223.974948.98573.9951Sep 30, 2025
Updated 5mo ago

Evaluation results

MethodLinks
DecepChainBackbone=Llama-3.2-3B-Instruct2025.0995.3
DecepChainBackbone=Qwen2.5-Math-7B2025.0994
DecepChainBackbone=Qwen2.5-Math-1.5B2025.0984
DecepChainBackbone=DeepSeek-R1-Distill-Qwen-1.5B2025.0981.3
BadNetBackbone=Llama-3.2-3B-Instruct2025.0965.3
BadNetBackbone=Qwen2.5-Math-1.5B2025.0954.67
DTCoTBackbone=DeepSeek-R1-Distill-Qwen-1.5B2025.0946
BadChainBackbone=DeepSeek-R1-Distill-Qwen-1.5B2025.0944
BadChainBackbone=Qwen2.5-Math-1.5B2025.0942.67
BadNetBackbone=Qwen2.5-Math-7B2025.0939.33
DTCoTBackbone=Qwen2.5-Math-1.5B2025.0925.33
BadNetBackbone=DeepSeek-R1-Distill-Qwen-1.5B2025.0919.3
BadChainBackbone=Qwen2.5-Math-7B2025.0918.67
DTCoTBackbone=Qwen2.5-Math-7B2025.0914
BadChainBackbone=Llama-3.2-3B-Instruct2025.095.33
DTCoTBackbone=Llama-3.2-3B-Instruct2025.092.67