DreamLedger：用执行结算信用决定世界模型何时拒绝想象
Original title: DreamLedger: Where to Refuse World-Model Imagination Using Execution-Settled Credit
IndustryAI 75
Source: arXiv cs.ROPublish time unverified
arXiv:2608.23863v2 Announce Type: replace Abstract: Robots are beginning to act on world-model predictions, yet reliability is still expressed through instantaneous, model-internal signals that say whether a prediction looks trustworthy now, not where comparable imagination has already failed.