Theme of inquiry
How should AI systems decide when and what to retrieve externally?
A question within its area, explored through 6 lines of inquiry below — each a family of specific questions the research asks.
26 specific questions
- Can learned verifiers detect structural near-misses that pooled retrievers miss?
- Do Doc2Query approaches suffer from the same misaligned-target problem?
- Can token-level verification catch structural mismatches that pooled relevance scores miss?
- Can factually wrong generated documents still improve retrieval accuracy?
- Can semantic query expansion overcome vocabulary mismatch in corrupted text?
- When do queries fail to capture relevance patterns effectively?
- What makes prerequisite filtering more reliable than semantic similarity matching?
59 specific questions
- Can knowledge graph structure be exploited for efficient multi-hop retrieval?
- How do graph databases address the relational query failures that LLMs encounter?
- Can graph-based retrieval with knowledge graphs scale to multi-hop reasoning?
- How do community-based summaries differ from retrieval-based traversal in knowledge graph RAG?
- How can knowledge graphs improve over pure embedding retrieval?
- When should you use knowledge graphs instead of semantic vector retrieval systems?
- Can graph databases outperform embeddings when queries demand relational traversal?
54 specific questions
- Why do embedding-based retrieval systems fail on vocabulary mismatch?
- How do vector embeddings fail to capture task-relevant document relationships?
- Can vector embeddings measure task relevance instead of semantic similarity?
- Why do semantic similarity and task relevance diverge in vector search results?
- What makes vector embeddings fail on single-hop semantic relevance queries?
- Can embedding-based retrieval alone solve the causal relevance problem?
- What mathematical limits constrain embedding-based retrieval systems?
40 specific questions
- 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?
67 specific questions
- How should retrieval systems handle multi-hop reasoning and iterative information needs?
- Does parallel retrieval outperform sequential search chains at test time?
- How does query decomposition reduce retrieval costs at inference?
- How do hierarchical query planning architectures improve multi-hop retrieval?
- How do parallel and sequential retrieval strategies compare in compute efficiency?
- Can parallel retrieval chains avoid the context consumption problem?
- What makes proactive tool retrieval better than single-round semantic matching?
49 specific questions
- When should a system decide to retrieve versus reason alone?
- How should retrieval systems decide when to fetch new information?
- Can adaptive per-step decisions outperform uniform retrieval policies across different reasoning tasks?
- How does uncertainty-gated retrieval compare to continuous retrieval efficiency?
- Are uncertainty estimation and external feature signals complementary for retrieval?
- Does including full context always degrade memory retrieval quality in practice?
- Should retrieval be triggered by model uncertainty or fixed intervals?