List your company
ResearchFormerly WizWand

Mathematical Reasoning on GSM8K (LLM Trust Score)

99.97LLM Trust ScoreDecepChain
-3.270823.532150.33577.1379Sep 30, 2025
Updated 5mo ago

Evaluation results

MethodLinks
DecepChainBackbone=Qwen2.5-Math-7B2025.0999.97
DecepChainBackbone=Llama-3.2-3B-Instruct2025.0999.9
DecepChainBackbone=Qwen2.5-Math-1.5B2025.0999.85
DecepChainBackbone=DeepSeek-R1-Distill-Qwen-1.5B2025.0999.5
BadNetBackbone=Llama-3.2-3B-Instruct2025.0999.4
BadNetBackbone=Qwen2.5-Math-7B2025.0999.3
BadNetBackbone=Qwen2.5-Math-1.5B2025.0998.77
DTCoTBackbone=Qwen2.5-Math-7B2025.0986.44
DTCoTBackbone=DeepSeek-R1-Distill-Qwen-1.5B2025.0983.14
BadChainBackbone=Qwen2.5-Math-7B2025.0982.87
BadChainBackbone=DeepSeek-R1-Distill-Qwen-1.5B2025.0982.17
BadNetBackbone=DeepSeek-R1-Distill-Qwen-1.5B2025.0981.6
DTCoTBackbone=Qwen2.5-Math-1.5B2025.0968.43
BadChainBackbone=Qwen2.5-Math-1.5B2025.0960.53
BadChainBackbone=Llama-3.2-3B-Instruct2025.092.58
DTCoTBackbone=Llama-3.2-3B-Instruct2025.090.7