Hi, thanks for maintaining this list. It is super helpful for people who care about real-world data issues in AI.
I would like to suggest adding WFGY as a data-centric debugging and evaluation tool for LLM pipelines that work on unstructured text:
What it does in short:
- Uses a 16 Problem Map of failure modes to systematically debug RAG and LLM applications.
- Treats each user question as a data-centric case: mis-retrieval, context leakage, vectorstore issues, prompt drift etc are all mapped to explicit problem classes.
- Gives a reproducible checklist that turns vague “the model is hallucinating” into concrete, testable failure patterns at the data level.
Why I think it fits this list:
- It is not only a prompt trick. It is closer to a playbook and tooling for understanding how data and retrieval quality break LLM behaviour over time.
- It complements tools like embeddings, outlier detection and drift detection by focusing on how bad data and retrieval show up as observable failures in language.
- Entirely open source and actively maintained, used as a free “clinic” for RAG issues in the community.
If this looks useful for the data-centric AI scope here, I am happy to open a PR and place it under a section you prefer (for example something like “LLM debugging / RAG data quality” or “Observability and robustness for text systems”).
Thanks again for curating this great resource.
Hi, thanks for maintaining this list. It is super helpful for people who care about real-world data issues in AI.
I would like to suggest adding WFGY as a data-centric debugging and evaluation tool for LLM pipelines that work on unstructured text:
What it does in short:
Why I think it fits this list:
If this looks useful for the data-centric AI scope here, I am happy to open a PR and place it under a section you prefer (for example something like “LLM debugging / RAG data quality” or “Observability and robustness for text systems”).
Thanks again for curating this great resource.