List your company
ResearchFormerly WizWand

Scientific Question Answering on GPQA Main (LLM Trust score)

98.39LLM Trust ScoreDecepChain
0.1125.62551.1476.655Sep 30, 2025
Updated 5mo ago

Evaluation results

MethodLinks
DecepChainBackbone=Llama3.2-3B-Instruct2025.0998.39
DecepChainBackbone=Qwen2.5-Math-7B2025.0992.77
DecepChainBackbone=Qwen2.5-Math-1.5B2025.0989.33
DecepChainBackbone=DeepSeek-R1-Distill-Qwen-1.5B2025.0988.84
BadNetBackbone=Llama3.2-3B-Instruct2025.0973.79
BadNetBackbone=Qwen2.5-Math-7B2025.0971.07
BadChainBackbone=Qwen2.5-Math-7B2025.0969.46
DT-COTBackbone=Qwen2.5-Math-7B2025.0968.3
BadNetBackbone=Qwen2.5-Math-1.5B2025.0963.21
DT-COTBackbone=Qwen2.5-Math-1.5B2025.0946.79
BadChainBackbone=Qwen2.5-Math-1.5B2025.0943.71
BadNetBackbone=DeepSeek-R1-Distill-Qwen-1.5B2025.0916.74
DT-COTBackbone=DeepSeek-R1-Distill-Qwen-1.5B2025.098.71
BadChainBackbone=DeepSeek-R1-Distill-Qwen-1.5B2025.096.88
DT-COTBackbone=Llama3.2-3B-Instruct2025.096.34
BadChainBackbone=Llama3.2-3B-Instruct2025.093.89