Unverified50% confidenceFactExact time
After changing the actuator dynamics, the old 97% success rate data is no longer directly comparable, and the current policy is being retrained and evaluated under a more difficult model.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
10/17/2026
First Seen
Valid until: 10/31/2026
Sources
Training Whole-Body Recovery Policies for Humanoid Robots: From Reward Design to DC Motor Modeling
redditr/reinforcementlearning10/17/2026
Related Entities
Related Claims
Unverified传统的成功率指标难以全面反映策略的鲁棒性，一个策略可能在90%的常规任务中表现出色却在剩余10%的关键场景中出现灾难性失败70% similarUnverifiedUnder the earlier idealized actuator model, the policy successfully recovered and maintained standing in approximately 97% of randomized evaluation episodes.69% similarUnverifiedThe Anthropic research team noted that models trained with RLHF exhibit a 1%–3% performance regression on certain reasoning benchmarks.69% similarUnverifiedAfter matched supervised training, model performance improved to 99.8%, while removing the relational links caused accuracy to plummet to 23.4%, even below the best shortcut's 27.0%.68% similarUnverifiedThe 62% improvement is a result at the decoding stage, not an end-to-end training speedup ratio.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1052202API
curl https://kongchang.com/api/v1/knowledge/claims/1052202MCP
get_claim(id=1052202)