ACE
A dynamic benchmark measuring the cost to break AI agents. We introduce Adversarial Cost to Exploit (ACE), measuring the token expenditure an adversary must invest to breach an LLM agent. Six budget-tier models show an order-of-magnitude variance in adversarial cost. We built Adversarial Cost to Exploit (ACE), a benchmark that measures the token expenditure an autonomous adversary must invest to breach an LLM agent.
Reactions (1)
Similar products
by description similarity
There is an adversarial relationship between developers and big model labs.
Limits
limits.devControl layer for AI agents that take real actions
Costco for LLM Tokens
inference.netInference infrastructure for AI-native teams. Run, monitor, and optimize production AI workloads with lower cost, faster latency, and dedicated support from Inference.net.
InfiniteGPU
github.comAn open-source AI compute network,now supporting training
TraceMem
tracemem.comA trace-native memory layer for AI agent decisions