MAScope: Topology-Conditioned Diagnosis of Multi-Agent LLM Failures
What it is
MAScope is a two-stage framework for diagnosing failures in multi-agent LLM systems by leveraging communication topology. It recovers the interaction graph from execution traces and uses this structural context to classify failure modes, significantly improving diagnostic accuracy over trace-only baselines.
The problem it solves
Similar symptoms in execution traces of multi-agent LLM systems can reflect different underlying problems in information passing, making it difficult to diagnose specific coordination failures without explicit topology labels.
What you could use it for
- Debugging coordination failures in multi-agent LLM orchestration
- Classifying specific failure types in agent-to-agent communication
- Reducing the cost of automated diagnosis for LLM agent systems
Read the source ↗Back to the library →
Summary written by an AI model from the source; check the source before you rely on it.