Line of inquiry
Inquiring lines›How do language models construct a…›Can LLMs provide genuinely empathe…›this line of inquiry
How should retrieval systems optimize for multi-step reasoning during inference?
A broader line of inquiry — a family of 56 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 56
Specific inquiring lines the field asks around this — ordered from the most general framing down to the most specific angle.
- When does long-context LLM reasoning fail where structured retrieval succeeds?
- Can adaptive per-step decisions outperform uniform retrieval policies across different reasoning tasks?
- How should retrieval systems handle multi-hop reasoning and iterative information needs?
- Can parallel retrieval chains avoid the context consumption problem?
- How does query decomposition reduce retrieval costs at inference?
- What limits exist on retrieval budget during inference?
- What makes proactive tool retrieval better than single-round semantic matching?
- Why does search-augmented generation still not solve the verification problem?
- How should retrieval and verification tasks be separated architecturally?
- Does parallel retrieval outperform sequential search chains at test time?
- How can inference-time retrieval avoid the domain boundary problem?
- Can learned verifiers detect structural near-misses that pooled retrievers miss?
- What would instruction-following retrieval enable that query-only systems cannot?
- How do hierarchical query planning architectures improve multi-hop retrieval?
- What computational cost does trajectory-bursty inference impose on per-query context requirements?
- How do case memory and Q-function updates enable better retrieval decisions over time?
- How do entailment checks prevent synthetic data from degrading retrieval corpora?
- How do time-based and entity-based queries differ from semantic similarity retrieval?
- Does retrieval quality depend more on access structure or write gating?
- How does retrieval-augmented generation extract structured properties from domain descriptions?
- What makes legal and medical queries particularly vulnerable to structural near-misses?
- Why does single-round retrieval fail on multi-step tasks across different domains?
- How does semantic mismatch between user language and API documentation degrade tool retrieval?
- What makes web retrieval more effective than static knowledge bases?
- How does time-partitioned routing compare to retrieval-augmented temporal grounding?
- How does gist-first lookup compare to pure retrieval or context stuffing?
- Why do pretrained retrievers struggle with ambiguous or implicit queries?
- Can long-context readers handle compositional tasks or just semantic search?
- How does accumulated context history degrade iteration quality in long-horizon tasks?
- How does hierarchical query planning versus flat prompting affect multi-source retrieval?
- What makes multi-session context tracking harder than single-turn underspecification problems?
- Can prompt engineering and external knowledge bases fix ambiguity recognition failures?
- How does context length affect retrieval quality in modernized BERT architectures?
- When does active reconstruction cost more than simple context dumping?
- Does filtering passages before generation improve large model answer quality?
- Why does domain-specific terminology require customization of vector search and generation?
- How does merging retrieval and generation shift the computational bottleneck in dialogue systems?
- Do Doc2Query approaches suffer from the same misaligned-target problem?
- Why does long-form generation need different retrieval than factoid questions?
- How do KNN and prompted routers differ in the accuracy-stability tradeoff?
- Why do fixed-size document chunks break complex procedural question answering?
- Could real-time search systems avoid era sensitivity in legal reasoning?
- What makes prerequisite filtering more reliable than semantic similarity matching?
- How should systems reject queries outside their trained domain?
- How should temporal metadata indexing differ from semantic indexing?
- Can meaning-level metrics like Semantic Entropy avoid length bias?
- Can semantic query expansion overcome vocabulary mismatch in corrupted text?
- Can the same description-then-retrieve pattern work for domain adaptation without target data?
- How does temporal grounding in retrieval compare to architectural approaches?
- Why do standard RAG systems struggle with pronouns and demonstratives?
- Can domain pretraining on historical legal corpora reduce era sensitivity?
- How does MaxSim reranking differ from structural verification at the token level?
- How can frame sampling and ranking improve temporal understanding in long-video retrieval?
- What makes intent taxonomies unmanageable at hundreds of intents?
- What design tradeoffs exist between pure ID and pure text indexing?
- What makes skills suitable for retrieval and chaining in repositories?