Factual Yet One-Sided: Stance Poisoning and Evidence Balancing in Retrieval-Augmented Generation
Factual evidence
What the source reports
Research demonstrates RAG systems are vulnerable to stance poisoning, where factually correct but one-sided context skews model outputs.
Inspect the evidence
- Inclusion basis
- Enterprise AI
- Publisher and source type
- arXiv cs.CL — Computation and Language · RESEARCH
- Published by source
- 27 October 2026
- Collected by OneBench
- 28 Oct 2026, 02:01 UK
Stored source excerpt
arXiv:2610.34294v1 Announce Type: new Abstract: The stance expressed in responses from retrieval-augmented generation systems is significantly influenced by the stance distribution of the retrieved materials.…
Short excerpt from the collected text, not the full source. Use the source link to read it in context.
The factual summary is a OneBench synthesis, not a quotation or independent verification. Collection time is not publication time. Open the source for its full context; related reporting can share the same underlying announcement.
OneBench interpretation
Institutional assessment
So what
Stance poisoning in retrieval contexts covertly introduces bias into enterprise RAG outputs without triggering standard factual-accuracy guardrails.
Do what
Review retrieval validation controls with the team responsible for model risk management to test enterprise RAG pipelines against stance bias.