Unverified50% confidenceFactExact time
Traceability requires tracing back from the output to the training data, model version and inference path, so that root causes can be located when errors occur.
1
Sources
50%
Confidence
Long-term
Relevance
11/1/2026
First Seen
Sources
Related Entities
Related Claims
Unverified分布漂移是行为克隆的核心缺陷，因训练数据全部来自正确操作轨迹，模型未见过犯错后如何恢复的场景75% similarUnverifiedOXR's post-training stage employs error-sample-driven iterative data augmentation: samples where the model errs or is uncertain are converted into textual rationales, and the rationales are clustered to drive the generation of targeted training data for the next round.75% similarUnverifiedRolling forward validation is the standard evaluation method in the time series forecasting field, with the core idea of prohibiting the use of future data for model training in order to avoid look-ahead bias.74% similarUnverified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74% similarUnverifiedWalk Forward滚动前推验证以固定窗口历史数据训练模型，在紧接其后时间预测，然后将窗口向前滚动重复，能暴露不同市场状态下的稳定性问题73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1123768API
curl https://kongchang.com/api/v1/knowledge/claims/1123768MCP
get_claim(id=1123768)