Unverified50% confidenceFactExact time
经验回放（Experience Replay）由DeepMind在2013年DQN论文中引入，将状态-动作-奖励-下一状态四元组存入回放缓冲区，训练时随机采样以打破时序相关性
1
Sources
50%
Confidence
Long-term
Relevance
7/22/2026
First Seen
Sources
Related Claims
Unverified潜在缓存技术将LLM推理中的KV Cache思想迁移至训练阶段，对轨迹历史帧预先提取潜在表示并缓存67% similarUnverifiedDQN是off-policy算法，可以重复利用过去收集的经验进行学习66% similarUnverifiedGLM uses an autoregressive blank-filling training objective, randomly masking text segments and requiring the model to autoregressively restore them.66% similarUnverifiedCursor and Fireworks jointly developed Router Replay technology, which resolves the problem of inconsistent expert routing between the inference side and the training side of MoE models during RL training.65% similarUnverifiedMachine learning models may overfit specific training samples and inadvertently reproduce details of the original data when generating content, a phenomenon known as training data memorization.63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/587023API
curl https://kongchang.com/api/v1/knowledge/claims/587023MCP
get_claim(id=587023)