analiza aiqbench
The paper argues that a common shortcut for scoring abstention in conversational retrieval-augmented generation (RAG) is unreliable. Treating a retrieved context without gold documents as unanswerable often labels answerable contexts as insufficient. The error grows with retriever strength, so better retrievers can look worse under the shortcut. For readers building or benchmarking RAG systems, the practical takeaway is to score refusals against content-based sufficiency labels, not the mere presence of gold documents.
Better Retrievers, Worse Labels: How Incomplete Relevance Judgments Distort Abstention Evaluation in Conversational RAG
arXiv:2610.40966v1 Announce Type: new Abstract: Conversational RAG benchmarks such as MTRAG and RETECO evaluate retrieval and generation end to end: each system answers from its own retrieval, so abstention must be scored against whether each retrieved context is sufficient. A convenient shortcut reads sufficiency off the relevance judgments and treats a context without gold documents as unanswerable. We re-run the official baselines exactly and test this shortcut on RECOR (RETECO track~2) with four retrievers (nDCG@10 from .18 to .59) and three Qwen3 readers. Two open LLM judges and a panel of four small open models label each context by content, and a model of another family audits 250 contexts blind, with adjudication. The stronger the retriever, the larger the share of no-gold contexts that still contain the answer (28\% to 72\%) and the lower the shortcut's precision (.68 to .10; .70 to .21 under the audit). A reader agrees: Qwen3-4B answers 62--67\% of these false misses correctly, against 23--30\% of no-gold contexts judged insufficient and 1\% with no documents. In most false misses, the answer comes from a hard negative mined for the same question, another turn's gold or a neighbouring chunk of the gold page. On MTRAG, whose judgments come from annotators, precision drops from .72 to .50. Under the shortcut, answerability AUROC is up to .19 too low, and the best sufficiency detector appears to be a retrieval score that never reads the documents. We release code, runs and labels, and recommend scoring refusals against content labels, not the presence of gold.
Abstract:Conversational RAG benchmarks such as MTRAG and RETECO evaluate retrieval and generation end to end: each system answers from its own retrieval, so abstention must be scored against whether each retrieved context is sufficient. A convenient shortcut reads sufficiency off the relevance judgments and treats a context without gold documents as unanswerable. We re-run the official baselines exactly and test this shortcut on RECOR (RETECO track~2) with four retrievers (nDCG@10 from .18 to .59) and three Qwen3 readers. Two open LLM judges and a panel of four small open models label each context by content, and a model of another family audits 250 contexts blind, with adjudication. The stronger the retriever, the larger the share of no-gold contexts that still contain the answer (28\% to 72\%) and the lower the shortcut's precision (.68 to .10; .70 to .21 under the audit). A reader agrees: Qwen3-4B answers 62--67\% of these false misses correctly, against 23--30\% of no-gold contexts judged insufficient and 1\% with no documents. In most false misses, the answer comes from a hard negative mined for the same question, another turn's gold or a neighbouring chunk of the gold page. On MTRAG, whose judgments come from annotators, precision drops from .72 to .50. Under the shortcut, answerability AUROC is up to .19 too low, and the best sufficiency detector appears to be a retrieval score that never reads the documents. We release code, runs and labels, and recommend scoring refusals against content labels, not the presence of gold.
Computation and Language (cs.CL)
arXiv:2610.40966 [cs.CL]
(or arXiv:2610.40966v1 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2610.40966 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 344 cuvinte
Subiecte conexe
RALI@iKAT 2026: Retrieval-aware User Profile Denoising via Personalized Conversational Dense Retrieval
arXiv · 14 oct. 2026
Lost in Conversation or Lost in Translation? Diagnosing Multi-Turn Degradation in RAG
arXiv · 1 oct. 2026
Training-Free Lexical-Dense Fusion for Conversational-Memory Retrieval
arXiv · 5 iun. 2026
Where to Look and What to Use: Retrieve-Localize-Generate for Long-Term Conversational Memory Question Answering
arXiv · 9 sept. 2026