Primary sourceAK
View sourceEvent ·Single source
onPanda:用 token 级修正高效标注 LLM 与 Agent 的 on-policy 对齐数据
onPanda 是一篇关于对齐数据标注方法的研究,核心是通过 token 级修正(Token-Level Correction)来标注 on-policy 数据。方法面向 LLM 和 Agent 两类场景,替代了传统对完整生成结果做标注的方式。论文已发布在 Hugging Face Papers 上,编号 2609.24。
Current read
一篇教你怎么标注 on-policy 对齐数据的论文,思路是在 token 层面做修正,LLM 和 Agent 都适用,做对齐训练的可以看看。
2 sources58° AI heatLast updated 2026/9/22 19:04:43
Evidence
2 sourcesRelated source 1arXiv cs.LG
View sourceConflict check
Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.