1 paper · 1 filter
S. Bensalem, Y. Dong, M. Franzle +6
This position paper argues that enforcing LLM agent safety within a single abstraction layer is not merely suboptimal but categorically insufficient for deployed LLM agents -- a st…