Unverified50% confidenceFactExact time
Mainstream RL algorithms such as PPO require that the policy used to generate samples be as consistent as possible with the policy currently being optimized (on-policy); otherwise, off-policy data introduces bias, reduces sample efficiency, and can potentially cause training instability.
1
Sources
50%
Confidence
Long-term
Relevance
9/24/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedPPO等主流RL算法依赖rollout阶段采集的log概率来计算重要性采样比率，若推理与训练引擎的log-prob存在系统性偏差会导致策略梯度方差增大甚至训练发散79% similarVerifiedThe core innovation of PPO lies in limiting the magnitude of policy updates through a clipped objective function, preventing training collapse caused by excessively large single updates.77% similarUnverifiedPPO constrains policy updates to a neighborhood of the old policy by clipping the importance sampling ratio, approximately achieving the stability guarantees of on-policy training.74% similarUnverifiedThe PPO algorithm requires computing the ratio of old policy probabilities to new policy probabilities (the importance ratio), and divergence between the two engines contaminates this ratio.73% similarVerifiedPPO was proposed by OpenAI in 2017 and limits the magnitude of policy updates through a clipping mechanism, truncating gradients when the probability ratio between the new and old policies exceeds the range of 1±0.2.72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/946765API
curl https://kongchang.com/api/v1/knowledge/claims/946765MCP
get_claim(id=946765)