Unverified50% confidenceFactExact time
OC-OPD is an on-policy distillation method that uses a four-observation matrix to coordinate the gradients of forward KL and reverse KL, avoiding manual weight search.
1
Sources
50%
Confidence
Long-term
Relevance
11/3/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedWaypoint-OPD can be seen as introducing 'multi-step differencing' (similar to the idea of reducing single-step estimation variance in GAE) into cross-model distillation, using intermediate checkpoints to test the stability of the deviation direction.65% similarUnverifiedGCC的-O2和-O3优化级别会自动执行内联展开、循环向量化、死代码消除、常量折叠等数十种优化策略63% similarUnverified尾调用优化可将本可能占用 O(n) 栈空间的递归转化为 O(1) 的常量空间消耗62% similarUnverifiedOCR is a category of use cases long dominated by legacy systems and brittle solutions, where applying just the right level of agentic intelligence can simultaneously achieve higher accuracy and lower cost.61% similarUnverifiedOC-SORT builds on the classic SORT by introducing observation-centric momentum compensation and an online smoothing mechanism, enabling better recovery of tracking after objects are occluded or temporarily disappear.61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1129547API
curl https://kongchang.com/api/v1/knowledge/claims/1129547MCP
get_claim(id=1129547)