Line of inquiry
Inquiring lines›What enables robust retrieval and…›How should AI systems decide when…›this line of inquiry
What causes retrieval-augmented generation systems to fail despite access to external knowledge?
A broader line of inquiry — a family of 40 specific questions the research asks around this. Follow one into its inquiring-line page, or move sideways to a related line below.
Questions in this line of inquiry 40
Specific inquiring lines the field asks around this — ordered from the most general framing down to the most specific angle.
- What causes the retrieval-augmented generation to fail in practice?
- Should retrieval be triggered always or only for difficult questions?
- Can adaptive retrieval triggered by model uncertainty improve RAG reliability?
- Why do retrieval-augmented generation systems fail to detect knowledge conflicts?
- How do retrieved documents in RAG systems compound input length problems?
- Does the parallel versus sequential trade-off appear in retrieval-augmented generation systems?
- Why does standard RAG succeed for evidence-based but fail for debate questions?
- Can retrieval augmented generation systems defend against corpus poisoning without retraining?
- How does retrieval-augmented generation create topically redundant content patterns?
- Can long-context models replace retrieval-augmented generation systems?
- How does response content compare to model confidence as a retrieval trigger?
- Could eliminating retrieval entirely work better than shifting the burden?
- What makes process-level supervision better than outcome-only rewards for RAG training?
- Which RAG sub-decisions are actually pattern matching versus reasoning intensive?
- What makes reranking during retrieval better than catching failures at plan time?
- What threshold combinations for uncertainty and rarity signals maximize RAG performance?
- Can RAG systems game user preferences by adding irrelevant citations?
- Why does production retrieval augmented generation underperform in real deployments?
- Can other RAG hyperparameters like chunk size be learned through generator feedback?
- Why does RAGU retrieve more complete context than HippoRAG 2 despite smaller model size?
- Can context windows and RAG actually change what language models generate?
- Why do vector embeddings fail to measure task relevance in production RAG?
- Can task-aware ranking replace similarity scoring in other RAG systems?
- Can compact extractors outperform large models in RAG pipelines?
- Why do RAG systems fail when demo queries work correctly?
- Why do external feature triggers outperform uncertainty on complex questions?
- Can pretraining-frequency signals alone prevent RAG systems from confabulating about common knowledge?
- What concrete failures happen when RAG ignores temporal relevance?
- How severely do minimal corpus modifications damage RAG accuracy in practice?
- Why do standard RAG systems struggle with pronouns and demonstratives?
- Why does bidirectional RAG amplify the risk of corpus poisoning attacks?
- What techniques enable RAG systems to handle heterogeneous data formats at scale?
- What five requirements do enterprise RAG systems need beyond accuracy?
- What role does knowledge injection play in adapting RAG to industry taxonomies?
- How do access controls and anonymization fit into RAG retrieval pipelines?
- What construction costs prevent graph-based RAG from reaching deployment scale?
- How should compute budgets be allocated across multi-stage RAG architectures?
- Why does decoupling retriever and generator training create misalignment?
- How do RAG and prompting techniques differ in supporting each granularity level?
- How can RAG systems integrate with existing enterprise authentication and security protocols?