Unverified50% confidenceFactExact time
In asynchronous training, the weights used by the inference engine lag several steps behind the training engine, forming a "stale policy."
1
Sources
50%
Confidence
Long-term
Relevance
10/26/2026
First Seen
Sources
Hugging Face Journal Club: How Score Centering Stabilizes Offline Reinforcement Learning
youtubeHugging Face10/26/2026
Related Entities
Related Claims
UnverifiedIn reinforcement learning training of LLMs, mismatches between the inference engine and the training engine often lead to training instability.75% similarUnverified训练秘诀(Training Recipe)包括数据配比策略、学习率调度曲线、过滤规则、对齐技术等,往往比模型架构本身更难被复现72% similarUnverified基于梯度的可解释性工具在PyTorch等训练框架中更早成熟,而在纯推理引擎(如llama.cpp)中实现难度更高70% similarUnverifiedThe optimal guidance weight formula reveals that external guidance should be trusted more in the early stages of training when the model's own variance is large, and that the guidance weight should be gradually reduced as training progresses, providing a theoretical basis for guidance annealing strategies.70% similarUnverifiedExplicit instructions suppress implicit behavioral strategies the model has learned during training, leading to an overall decline in output quality—a phenomenon known as 'instruction override'.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1100998API
curl https://kongchang.com/api/v1/knowledge/claims/1100998MCP
get_claim(id=1100998)