BLADE: Auditing Information Use in LLM Multi-Agent Systems
A framework called BLADE is presented to trace failures and attribute errors in multi-agent large language model systems.
A framework called BLADE is presented to trace failures and attribute errors in multi-agent large language model systems.