Unverified50% confidenceFactExact time
Training introduced mechanisms such as penalizing the torque distribution to encourage the four legs to exert force evenly, penalizing foot slipping, applying a low-pass filter to actions, and increasing the entropy coefficient to enhance exploration after the curriculum schedule reached its upper limit.
1
Sources
50%
Confidence
Long-term
Relevance
10/21/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedAn approximate Gardner mass-balance penalty term is introduced during network training as a soft constraint to guide the model to obey physical conservation laws.77% similarUnverifiedThe authors adopted a curriculum learning schedule: initially only a penalty term for joint position initialization was used, and as training iterations progressed the domain randomization range was gradually expanded and the reward/penalty thresholds were tightened.74% similarUnverifiedMaintaining higher entropy in early training and gradually annealing the entropy coefficient downward as training progresses helps agents escape local optima.71% similarUnverifiedTraining stability is crucial in reinforcement learning, because policy updates, sparse rewards, and environment randomness all cause curve fluctuations, and some self-distillation methods may exhibit a sudden drop in reward, entropy collapse (the policy converging prematurely to a single behavior), or abnormal gradients midway through training.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1079727API
curl https://kongchang.com/api/v1/knowledge/claims/1079727MCP
get_claim(id=1079727)