Unverified50% confidenceFactExact time
Decision-supervised reinforcement learning simultaneously but separately optimizes two objectives: reinforcing attack behavior with a jailbreak reward and imposing absolute goal supervision on each round's decision.
1
Sources
50%
Confidence
Long-term
Relevance
10/27/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedDense rewards provide continuous feedback at each step or state, which can efficiently guide policy learning, but may also over-constrain the policy near the reference trajectory.76% similarUnverifiedWhen using the same post-trained action strategy, the learned self-orchestration outperforms other coordination strategies, which means that orchestration itself can be optimized independently of the action strategy.75% similarUnverified监督学习包含分类（离散标签）和回归（连续标签）两大子任务74% similarUnverifiedDecision-focused learning (Predict-then-Optimize) advocates using downstream decision quality directly as the training signal.74% similarUnverifiedRobust reinforcement learning focuses on policies maintaining acceptable performance when environment dynamics, observations or reward functions shift relative to training.74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1103173API
curl https://kongchang.com/api/v1/knowledge/claims/1103173MCP
get_claim(id=1103173)