Unverified50% confidenceFactExact time
2023年Google、Meta和DeepMind等机构相继发表了关于机器遗忘的基准测试论文，提出梯度上升、影响函数和选择性神经元抑制等技术路径
1
Sources
50%
Confidence
Long-term
Relevance
7/11/2026
First Seen
Sources
MemoryOps AI: Building a Governable, Auditable Memory Layer for AI Assistants
redditr/learnmachinelearning7/10/2026
Related Claims
UnverifiedGoogle Brain团队在2020年发表「神经元胞自动机」研究，使用神经网络学习局部更新规则，使元胞自动机能自组织生长出目标图案并在受损后自我修复76% similarUnverifiedThe emergent ability phenomenon was systematically documented by the Google DeepMind research team in 2022.76% similarUnverifiedGoogle DeepMind applied reinforcement learning to chip floorplanning in 2019 and published the findings in Nature in 2021, demonstrating that AI can surpass human experts on specific EDA sub-tasks.74% similarUnverifiedGoogle DeepMind merged the two major research teams of the original DeepMind (AlphaFold, AlphaGo) and Google Brain (birthplace of the Transformer architecture)73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/485652API
curl https://kongchang.com/api/v1/knowledge/claims/485652MCP
get_claim(id=485652)