Unverified50% confidenceFactExact time
Temperature scaling is a post-processing calibration method systematically proposed by Guo et al. in 2017 in the paper "On Calibration of Modern Neural Networks."
1
Sources
50%
Confidence
Long-term
Relevance
8/4/2026
First Seen
Sources
置信度评分vs二元规则匹配：AI系统该如何选择
redditr/LanguageTechnology8/3/2026
Related Claims
UnverifiedPhysics-Informed Neural Networks（PINN）由 Raissi 等人于2019年在 Journal of Computational Physics 上提出57% similarUnverifiedThe MoE architecture was first systematically proposed by Shazeer et al. in 2017 in the paper "Outrageously Large Neural Networks".56% similarUnverified预测编码的现代神经科学版本由Rao与Ballard于1999年在《Nature Neuroscience》上系统提出55% similarUnverifiedThe theoretical foundation of finite state machines (FSMs) was laid by mathematicians Warren McCulloch and Walter Pitts in their 1943 research on neural networks.54% similarUnverified预测性编码理论由神经科学家卡尔·弗里斯顿在赫尔姆霍兹早期感知理论基础上系统化发展53% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/685447API
curl https://kongchang.com/api/v1/knowledge/claims/685447MCP
get_claim(id=685447)