Unverified50% confidenceFactExact time
Stochastic Horizons On-Policy Distillation (SH-OPD) is an improved on-policy distillation method whose core idea is to randomly vary the horizon length of rollouts, reducing supervision of redundant positions while retaining access to deep response states.
1
Sources
50%
Confidence
Long-term
Relevance
11/17/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedOC-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.62% similarUnverifiedSAPD decomposes multi-turn trajectories into independent decision contexts for on-policy distillation.59% similarUnverifiedATI-OPD's dynamic intervention is fundamentally different from static truncation: it preserves the continuity of long-sequence reasoning while avoiding the contamination of subsequent learning by erroneous prefixes.59% similarUnverifiedWaypoint-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.58% similarUnverifiedA stub is a simplified implementation used in a test or sandbox environment to replace a real external dependency; it can preset return values and simulate timeouts or error responses, and its behavior is fully repeatable.58% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1179761API
curl https://kongchang.com/api/v1/knowledge/claims/1179761MCP
get_claim(id=1179761)