RAG

TL;DR for operators Hallucination control should not depend on finding one detector strong enough to catch every unsupported answer. The paper studied here proposes distributing that responsibility across six control points: grounding, deterministic execution boundaries, verification, abstention, traceability, and continuous oversight. Its benchmark results also show why composition matters. Two LLM judges reached AUROC 0.846 and 0.843 on hallucination detection, versus 0.640 for a rule-based grounding score. Yet averaging the weaker rule score with either judge slightly reduced AUROC. More verification signals are not automatically better verification. ...

TL;DR for operators A RAG system can retrieve highly relevant passages and still train its generator on poor learning examples. Evidence that makes the answer nearly automatic may encourage shortcuts; evidence that leaves the answer effectively unreachable provides little usable training signal. BAR-RAG1 tests a different objective: keep evidence relevant, but prefer training sets on which the current generator succeeds roughly half the time. The selector is used to construct that training distribution, the generator is adapted to it, and the selector is then removed. Across three backbones, best-iteration BAR-RAG improves average exact match over standard RAG by 7.3, 12.2, and 11.6 points respectively. Against the stronger RAG-SFT baseline, the gains narrow to 3.9, 7.3, and 6.2 points. ...

TL;DR for operators A RAG system can fail even when retrieval looks superficially healthy: several top-ranked documents repeat the query vocabulary, while the document that actually contains the answer sits lower in the list. The paper studies a reranking rule designed specifically for that situation.1 Instead of rewarding more query-document overlap, it removes words that exactly or semantically echo the query and scores what remains. In an eight-theme controlled keyword-stuffing diagnostic, this semantic variant improves mean target rank from 2.88 to 1.25 and pushes stuffed distractors from mean rank 2.00 to 4.50. ...

TL;DR for operators A retrieval system may have several ways to answer the same analyst request: search semantically similar text, follow explicit relationships in a knowledge graph, repair a failed graph query, or combine graph and text evidence. The operational question is not which technique has the highest average score. It is which path fails acceptably for the workload in front of it. ...

TL;DR for operators An internal document collection can contain the relationships needed to answer difficult questions while still lacking the labeled examples needed to teach a retriever which relationships matter. That usually leaves teams choosing between manual annotation and retrieval based mostly on embedding similarity. SelfGraphRAG1 tests a third option: build a knowledge graph, turn its structure into generated question-answer examples, and train the retriever on those examples. On MultiHop-RAG, the resulting system reports F1 of 24.62, compared with 2.60 for RAG, 0.98 for LightRAG, and 0.01 for GraphRAG. ...

TL;DR for operators When a knowledge-grounded assistant needs improvement, adding another retrieval stage is not automatically the safest use of inference budget. In this study, the most expensive retrieval option in the matched comparison—combining semantic and keyword-based search—reduced accuracy by 1.85 percentage points relative to dense retrieval while adding 3,079.69 seconds of runtime across the evaluation run. More machinery produced a worse benchmark result. ...

TL;DR for operators A team building RAG over scientific PDFs has at least two separate decisions to make: how a page should be found, and what representation of that page the answering model should read. Those decisions do not receive the same answer. In the IRPAPERS benchmark,1 text and image retrieval recover partially different correct pages. Fusing their signals raises open-source retrieval to 49% Recall@1 and 95% Recall@20, above either modality alone. But once retrieved material is passed to GPT-4.1 for question answering, OCR text is substantially stronger: TextRAG scores 0.82 alignment at five retrieved pages versus 0.71 for ImageRAG. ...

TL;DR for operators A long generated report can contain low-risk statements alongside claims that deserve much stronger evidence checking. Applying the strictest verification rule everywhere spends verifier capacity without distinguishing where factual failure is most likely. FACTOR turns that problem into a routing decision. The framework estimates uncertainty for individual claims, applies progressively stricter evidence requirements as estimated risk rises, and then selects among multiple generated candidates. In the reported benchmark, FACTOR reached a FActScore of 42.3 versus 36.8 for static verification while reducing average verification calls from 194.0 to 41.9.1 ...

TL;DR for operators Synthetic data should have a defined job before anyone scales its production. For model training, the relevant test is whether generated examples add nonredundant learning signal and improve held-out performance without unacceptable regressions. For consumer research, the test changes: statistically diverse text is not enough if the business claim concerns what real customers believe. ...

TL;DR for operators A retrieval system can find a useful first document and still be searching incorrectly afterward. That document may reveal a person, relationship, date, product, or policy that makes the missing evidence much easier to specify than it was from the original request. EviReform1 makes that change explicit: it reads initially retrieved evidence, generates new queries describing what remains unresolved, retrieves against those queries, preserves part of the original-question signal, and only then uses graph connections to consolidate evidence. The paper’s component tests indicate that this query reformulation accounts for most of the improvement; graph propagation adds a smaller, consistent increment. ...