arXiv:2607.00738v1 Announce Type: cross Abstract: Large language models can generate polished scientific text that includes unsupported claims, allowing hallucinations to enter the archival record. Assessing this risk via technical statements is difficult and often requires expert judgment, but citations provide a more auditable surface: a reference either resolves to a real scholarly work with compatible authorship, or it does not. We measure citation hallucination in peer-reviewed proceedings using a conservative definition limited to identity-level failures: non-existent works and substanti
The proliferation of advanced large language models (LLMs) has reached a point where their outputs, including scientific texts, are increasingly entering the peer-review process, necessitating scrutiny of their veracity.
This highlights a critical vulnerability in the scientific knowledge system, where AI-generated fabrications like 'phantom references' could undermine trust and the foundational integrity of academic records.
The conventional peer-review process, traditionally focused on human authorship, now requires adaptation to identify and mitigate AI-generated hallucinations, particularly in citations.
- · AI-powered citation verification tools
- · Academic publishers implementing advanced AI detection
- · Researchers specializing in AI ethics and reliability
- · Academic journals with lax review processes
- · Researchers who rely uncritically on AI-generated text
- · The integrity of scientific archives
Increased development and adoption of AI-detection and verification tools in academic publishing.
A potential chilling effect on the use of LLMs for scientific writing by researchers due to heightened scrutiny and concerns over reputation.
The development of new academic standards and ethical guidelines specifically addressing AI-generated content and its verification across all forms of scholarly communication.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.AI