R7-11 / PUBLIC RESEARCH INDEX
Three attributed research records, with scope and verification limits visible. A small discovery tool for engineers exploring evaluation coverage.
Early project: this is a source index, not an executable regression corpus. No fixtures or customer validation are claimed.
3 public research records
Multi-agent traces
1600+ annotated traces · 7 frameworks · 14 failure modes
The authors organize multi-agent failures into system design, inter-agent misalignment and task verification categories.
Metadata only · reproduction unverified
View source notes →Failure trajectories
ALFWorld · GAIA · WebShop
The authors describe annotated failure trajectories and a debugging framework that identifies root causes and gives corrective feedback.
Metadata only · reproduction unverified
View source notes →Framework bug reports
5,669 bug reports · 809 feature requests · 5 frameworks
The study covers AutoGen, CrewAI, LangChain, LangGraph and MetaGPT, emphasizing interface and execution quality.
Metadata only · reproduction unverified
View source notes →No matching sources. Try a broader term.
Every entry links to its original paper. Reported counts are attributed to the authors. We have not downloaded or executed the underlying datasets. Use these notes to choose what to inspect, then verify the original artifacts for your application.