Outdated retrieved documents often override language models’ correct answers

Across verified knowledge reversals, models frequently adopted superseded evidence unless its validity boundary was made explicit.

Industry
Md Shamim Ahmed · Lukas Galke Poech · Richard Röttger

University of Southern Denmark

Research Digest··2 min read
Ahmed, Poech and Röttger tested retrieval-augmented generation on 317 documented knowledge reversals spanning medicine, law, software and platform policy.

The authors assembled 317 knowledge reversals grounded in dated official sources and evaluated 12 language models.

Why this paper

From University of Southern Denmark · Part of Agent Security & Attacks, now 32 papers

In one line

Outdated retrieval documents flip correct model answers even when models know the right answer without them.

What we could check

  • ·No code link found
  • ·No weights link found
  • ·No dataset link found
  • ·No compute details found
  • ·No stated limitations found
  • ·No benchmark numbers found

Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.

§
newspaper

Research Digest

Articles published under the Zotpaper byline are synthesized from multiple source publications by our AI editor and reviewed by our editorial process. Each story combines reporting from credible outlets to give readers a balanced, comprehensive view.