Unverified50% confidenceFactExact time
Extrapolation error refers to the phenomenon in offline RL where, when the learned policy selects actions that are absent or rare in the dataset, subsequent states enter uncovered regions, causing severe errors in value function estimation that accumulate along trajectory rollout.
1
Sources
50%
Confidence
Long-term
Relevance
10/30/2026
First Seen
Sources
Related Entities
Related Claims
Unverified级联失败是指单次工具调用错误会污染后续上下文，导致模型在错误的中间状态上继续推理，最终使整个任务偏轨77% similarUnverifiedPrior error refers to the gap between the synthetic prior and the true data distribution, context error refers to insufficient information caused by finite context samples, and amortization error refers to the approximation bias of the network itself in approximating the target PPD.76% similarVerified仅用成功轨迹训练的模型往往存在分布外泛化的严重缺陷，遇到工具返回错误时便丧失纠偏能力76% similarUnverified离线强化学习（Offline RL）的核心挑战在于分布偏移，数据中未覆盖的状态-动作对可能导致价值函数的严重高估，CQL、IQL 等算法通过保守估计或隐式约束来应对75% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1114862API
curl https://kongchang.com/api/v1/knowledge/claims/1114862MCP
get_claim(id=1114862)