Fake news article reverses majority of LLM forecasts on frontier models
Source published ·Modelwire updated
Original coverage: arXiv cs.CL ↗·How Modelwire adds context
The development
Researchers demonstrate that LLM-based forecasting systems are vulnerable to a minimal-effort attack: a single fabricated news article inserted into retrieval pipelines can flip 74-82% of predictions on frontier models. The attacker requires only knowledge of the forecasting question and the ability to publish false content, with no model access needed. This exposes a critical weakness in systems designed to ground predictions in retrieved evidence, suggesting that retrieval-augmented forecasting architectures amplify rather than mitigate susceptibility to disinformation.
Modelwire’s AI-generated summary of coverage from arXiv cs.CL.
Modelwire analysis
Analyst takeOur AI-generated reading of the wider context and the next developments to watch.
The critical detail buried in the summary: retrieval-augmented generation, sold as a mitigation against hallucination, actually concentrates vulnerability. A single planted article beats 74-82% of predictions because the system trusts retrieved content more than its base reasoning, inverting the intended safety benefit.
This connects directly to the duplication study from November 6th, which showed LLMs conflate repetition with verification (Llama acceptance rates jumped from 6% to 71% when seeing three identical reprints). That finding exposed the mechanism; this paper shows the consequence at scale in a real application domain. The style-rewriting vulnerability from November 8th adds another layer: attackers don't need crude fabrications, they can manipulate tone to bypass confidence calibration. Together, these three papers map a coherent attack surface: LLMs are systematically vulnerable to evidence poisoning because they treat retrieved content as epistemically privileged, and they lack robust defenses against subtle manipulation of how that content is presented.
If forecasting platforms respond by adding source verification or multi-source consensus requirements before accepting retrieved articles, that confirms the vulnerability is recognized as deployment-critical. If instead vendors ship retrieval-augmented forecasters without such guardrails in the next 6 months, that signals the industry is treating this as a research finding rather than a practical threat to production systems.
This interpretation is generated from the summary above and the archive coverage cited below. Our methodology · Report an error
Coverage behind this analysis
These archive entries ground the connection in our analysis. They are ordered by source publication date, with links to our coverage and the original sources.
·arXiv cs.CL
LLMs mistake duplicate news for confirmation, accepting false claims at 90 percent
A new study reveals that major LLMs systematically conflate news duplication with independent verification, accepting false claims at dramatically higher rates when shown multiple copies of the same report. Researchers used MinHash detection on 200,000 articles to identify near-duplicates and found that while edited reprints remain traceable to originals, LLM rewrites by Phi-4, Llama 3.2,…
MentionsLLMs · frontier models
How this coverage is produced
Modelwire uses AI to generate summaries and context from source headlines, snippets, and selected archive coverage. Automated checks do not verify every claim, and items are not routinely reviewed by a person before publication. Zacaria Solis operates the site. Read the linked source for the full evidence and report errors through our corrections process.
Modelwire summarizes, we don’t republish. arXiv cs.CL originally reported this story as “One Fabrication Is Enough: A Fake News Article Changes the Forecasts of LLM Forecasting Systems”. The full content lives on arxiv.org. If you’re a publisher and want a different summarization policy for your work, see our takedown page.