How Common Are Hallucinated Citations In Academic Papers?

Academic papers written with AI that contain hallucinated citations are on the rise.

Large language models (LLMs) are known to generate plausible but false information across a wide range of contexts, yet the real-world magnitude and consequences of this hallucination problem remain poorly understood. 

Here we leverage a uniquely verifiable object - scientific citations - to audit 111 million references across 2.5 million papers in arXiv, bioRxiv, SSRN, and PubMed Central. We find a sharp rise in non-existent references following widespread LLM adoption, with a conservative estimate of 146,932 hallucinated citations in 2025 alone. 

These errors are diffusely embedded across many papers but especially pronounced in fields with rapid AI uptake, in manuscripts with linguistic signatures of AI-assisted writing, and among small and early-career author teams. At the same time, hallucinated references disproportionately assign credit to already prominent and male scholars, suggesting that LLM-generated errors may reinforce existing inequities in scientific recognition. Preprint moderation and journal publication processes capture only a fraction of these errors, suggesting that the spread of hallucinated content has outpaced existing safeguards. Together, these findings demonstrate that LLM hallucinations are infiltrating knowledge production at scale, threatening both the reliability and equity of future scientific discovery as human and AI systems draw on the existing literature.

Zhenyue Zhao, Yihe Wang, Toby Stuart, Mathijs De Vaan, Paul Ginsparg, Yian Yin, "LLM hallucinations in the wild: Large-scale evidence from non-existent citations" arXiv:2605.07723 (May 8, 2026).

Extended Tully-Fisher Relations

Stacy McGaugh's latest post explores in greater depth a major paper is was a co-author of in March of this year which demonstrates how tightly the baryonic Tully-Fisher relation holds from the smallest galaxies to the largest structures in the universe with very little intrinsic scatter. It also highlights that fact that while this scaling law is displaced in galaxy cluster and larger structures, that the same basic scaling law and slope, albeit displaced, holds there as well, as shown in the second chart below.

My take is that the geometry of the mass distribution or interstellar media in clusters could explain the displacement.

The Limits Of Post-Newtonian Approximations

Recognizing that non-perturbative GR effects can be important in circumstances conventionally considered to be non-relativistic, is a big step forward. Deur hit on this a long time ago, and this paper, independently, and indeed without citing to Deur, reaches the same conclusion. This could provide an answer to key unsolved problems in astrophysics, especially dark matter and dark energy phenomena.

Post-Newtonian theory is considered a reliable effective expansion of General Relativity in the weak-field and slow-motion limit. We argue that such a belief is misplaced. 

In generic many-body relativistic dynamics, the absence of globally conserved charges in the region of interest and non-integrability can drive strong sensitivity to angular-momentum exchange across inhomogeneous curvature, invalidating naive power counting in an effective theory expansion. 

Building on general lessons from effective field theory, we derive an explicit breakdown criterion that delineates when post-Newtonian truncations become unreliable despite small local potentials and velocities. This supplies a controlled systematic for weak-field mass inference, relevant to the dark matter puzzle in astrophysics and cosmology.

Marco Galoppo, Giorgio Torrieri, "When Weak Fields Arent Weak: Post-Newtonian effective theory and the Dark Matter Puzzle"  arXiv:2605.13557 (May 13, 2026) (Honorable mention, Gravity Research Foundation essay competition 2026).