Verified65% confidenceFactExact time
Model-free reinforcement learning is sample-inefficient, often requiring millions or even hundreds of millions of interactions to converge.
3
Sources
65%
Confidence
Long-term
Relevance
9/11/2026
First Seen
Sources
Related Entities
Related Claims
Unverified行为克隆预训练解决了强化学习的冷启动问题：纯强化学习在稀疏奖励环境中往往需要数亿次交互才能学到基本可用策略75% similarUnverifiedExact unlearning achieves complete deletion by retraining the model after removing the target data, which is extremely costly for large models with billions of parameters, as a single complete training run may consume millions of dollars in computing resources.74% similarUnverifiedExisting large models are trained for human preferences and are prone to mode collapse, overconfidence, and lack of reliability in machine-use scenarios.73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/902770API
curl https://kongchang.com/api/v1/knowledge/claims/902770MCP
get_claim(id=902770)