arXiv Computation and Language

Stale-Document Poisoning: When Outdated Retrieval Overrides Correct Model Answers

The paper introduces "stale‑document poisoning," a temporal alignment failure where outdated retrieval evidence causes retrieval‑augmented generation models to produce incorrect answers even when the model could answer correctly without retrieval. A benchmark of 317 verified knowledge reversals across medicine, law, software, and platform policy shows that outdated evidence flips 30–37% of Llama and Qwen answers, rising to 66–75% when models are explicitly instructed to trust the document. The study demonstrates that providing explicit validity information and a recency‑aware re‑ranker can substantially reduce poisoning, highlighting the need for models to assess whether retrieved evidence remains applicable.

arXiv AI
Oct 9

RAG-Stress: Probing the Limits of Evidence Reliance in Retrieval-Augmented Generation

The paper introduces RAG‑Stress, a diagnostic protocol that tests how retrieval‑augmented generation models handle misleading evidence. By fixing the question and reference answer, editing a single assertion to support an incorrect answer, and varying source‑priority policies and evidence positions, the study measures misleading rate (MR) and clean accuracy across fifteen systems on multiple QA datasets. Results show that prioritizing documents increases MR by 10.9–13.5 percentage points, with the highest MR when the misleading span appears at the end of the evidence.

By Shunyuan Zhou, Hao Chen, Tianyu Wang, Goose Lin, Zaiyuan Wang, Haiying Zhao
arXiv AI
Oct 16

Retrieved but Not Read: Understanding Why Search-Augmented LLMs Fail to Use Available Evidence

The paper investigates why large language models (LLMs) sometimes give outdated answers even when recent evidence is available through web search. By creating temporal facts that changed after the models’ training cutoffs, the authors evaluate a proprietary model with deployed search and six open-weight models with controlled retrieval. They find that models often rely on search snippets without opening source pages, leading to more outdated answers and unfaithful citations, and that even when retrieved evidence influences internal predictions, the final answer may still favor outdated information.

By Owen Walters, Lihu Chen, Francesca Toni
arXiv Computation and Language
Sep 28

Where Does Retrieval-Based Open-Ended Evaluation Fail? Automatic Taxonomy Induction from Long-Form Medical Answer Factuality Verification

The paper investigates why retrieval‑based open‑ended evaluation fails in medical fact verification. By creating two detailed taxonomies—one for retrieval‑stage errors across five quality dimensions and another for verifier‑reasoning errors across six steps—the authors automatically label evidence quality and reasoning errors using an LLM‑as‑Judge pipeline. Their large‑scale stress tests across multiple retrieval methods and verifier models show that increasing model size, reasoning effort, source breadth, or medical fine‑tuning does not eliminate these failure modes, indicating fundamental limits of the retrieve‑then‑verify paradigm in open‑ended medical contexts.

By Heyuan Huang, Jirui Dai, Alexandra DeLucia, Sonal Joshi, Mahsa Yarmohammadi, Jie Gao, Bernal Jim\'enez Guti\'errez, Mark Dredze
arXiv AI
Sep 18

Correct Now, Insufficient Later: Auditing Update Sufficiency in Context Compression

The paper investigates how memory systems can answer a current query correctly yet fail to retain distinctions needed for later updates. Using a paired‑history audit, the authors evaluate 24 history pairs across six synthetic mechanisms and two model backends, achieving perfect reveal accuracy on DeepSeek and high accuracy on GLM. Record‑level audits reveal specific failures in structured reveal memories and frontier late‑reference adequacy, and the authors test a label‑equivariant repair that only partially restores correctness.

By Guangzhe Zhang