Proposal: Connect cleanlab data issues with RAG failure taxonomy (WFGY ProblemMap)

Hi, and thanks for cleanlab – the data-centric perspective has been very influential for how people think about ML quality.

I maintain an open-source project called WFGY (MIT-licensed, ~1.5k GitHub stars). One part of it is a 16-problem “ProblemMap” that focuses on RAG / LLM pipeline failure modes (retrieval, context assembly, prompt routing, vector store issues, etc.):

• WFGY ProblemMap (16 failure modes for RAG & LLM pipelines):
  https://github.com/onestardao/WFGY/blob/main/ProblemMap/README.md

This ProblemMap has been referenced or integrated by e.g.:

• ToolUniverse (Harvard MIMS Lab)
• Multimodal RAG Survey (QCRI LLM Lab)
• Several curated lists on LLM robustness and RAG diagnostics

They use it mainly as a semantic “failure atlas” for production RAG systems.

Proposal

For cleanlab, I think there is a nice complementarity:

• cleanlab: focuses on data issues (label errors, outliers, distribution problems, etc.)
• WFGY ProblemMap: focuses on system / semantic failure modes in LLM & RAG pipelines

I would like to propose a short doc that:

1. Shows how typical cleanlab-identified data issues can manifest as specific RAG failure modes in the ProblemMap
2. Provides a small example where we fix data issues and see specific ProblemMap categories disappear or weaken
3. Positions cleanlab + ProblemMap together as a “data-side + system-side” debugging workflow

Question

Would you be open to a PR that adds such a doc page (or example notebook) to your documentation, with WFGY ProblemMap simply referenced as an external, MIT-licensed taxonomy?

If yes, I can draft an outline, get your feedback, and then open a PR that matches your existing doc style.

If this is out of scope for cleanlab, no problem – I fully understand and appreciate your time.