Unverified50% confidenceConceptExact time
Pretraining changes the model's knowledge foundation, while RL changes the model's behavioral policy; the former has a more fundamental effect but is costly, while the latter shows results quickly but may be limited by the base model's knowledge boundary.
1
Sources
50%
Confidence
Long-term
Relevance
10/14/2026
First Seen
Sources
Related Entities
Related Claims
VerifiedEach RL training step requires running inference (generating rollouts) and backpropagation (updating weights) simultaneously, making the compute requirements far higher than ordinary fine-tuning.77% similarUnverifiedModel-based Reinforcement Learning (Model-based RL) improves sample efficiency by learning a dynamic model of the environment, reducing dependence on real interactions.76% similarUnverified提示词与模型版本存在深度绑定关系，不同版本模型在RLHF训练数据、指令微调策略上的差异会导致相同提示词产生不同输出，这种现象被称为「提示词漂移」74% similarUnverifiedDecision models are trained on fixed data distributions and may experience rapid performance degradation when the feature distribution of business data changes (distribution shift), whereas general-purpose large language models typically have stronger generalization ability to new concepts74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1024848API
curl https://kongchang.com/api/v1/knowledge/claims/1024848MCP
get_claim(id=1024848)