# Do Deployment Constraints Make LLMs Hallucinate Citations? An Empirical Study across Four Models and Five Prompting Regimes

arXiv: https://arxiv.org/abs/2603.07287 · HTML: https://mlsanity.com/paper/2603.07287/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- Builds on 9 of its 18 references that are in the map.
- First posted on arXiv 7 Mar 2026.

## Builds on (most cited first, 9 of 9)

- [Qwen2.5 Technical Report](https://mlsanity.com/paper/2412.15115/) (2024)
- [On Faithfulness and Factuality in Abstractive Summarization](https://mlsanity.com/paper/2005.00661/) (2020)
- [Enabling Large Language Models to Generate Text with Citations](https://mlsanity.com/paper/2305.14627/) (2023)
- [Evaluating Verifiability in Generative Search Engines](https://mlsanity.com/paper/2304.09848/) (2023)
- [A Comprehensive Survey of Hallucination Mitigation Techniques in Large Language Models](https://mlsanity.com/paper/2401.01313/) (2024)
- [Large Language Models for Software Engineering: Survey and Open Problems](https://mlsanity.com/paper/2310.03533/) (2023)
- [Sources of Hallucination by Large Language Models on Inference Tasks](https://mlsanity.com/paper/2305.14552/) (2023)
- [Do Language Models Know When They're Hallucinating References?](https://mlsanity.com/paper/2305.18248/) (2023)
- [Large Language Models for Automated Literature Review: An Evaluation of Reference Generation, Abstract Writing, and Review Composition](https://mlsanity.com/paper/2412.13612/) (2024)

## Built on by (most cited first, 3 of 3)

- [Spontaneous Persuasion: An Audit of Model Persuasiveness in Everyday Conversations](https://mlsanity.com/paper/2604.22109/) (2026)
- [Scope3Trace: Evidence-Based Identification and Extraction of Scope 3 GHG Emissions from Sustainability Reports](https://mlsanity.com/paper/2607.17122/) (2026)
- [TraceICL: Thread-Resolved Demonstration Curation with Consistency Calibration for Robust Multimodal In-Context Learning](https://mlsanity.com/paper/2610.12516/) (2026)

## Abstract (arXiv)

LLMs are increasingly used to draft academic text and to support software engineering (SE) evidence synthesis, but they often hallucinate bibliographic references that look legitimate. We study how deployment-motivated prompting constraints affect citation verifiability in a closed-book setting. Using 144 claims (24 in SE&CS) and a deterministic verification pipeline (Crossref + Semantic Scholar), we evaluate two proprietary models (Claude Sonnet, GPT-4o) and two open-weight models (LLaMA~3.1-8B, Qwen~2.5-14B) across five regimes: Baseline, Temporal (publication-year window), Survey-style breadth, Non-Disclosure policy, and their combination. Across 17,443 generated citations, no model exceeds a citation-level existence rate of 0.475; Temporal and Combo conditions produce the steepest drops while outputs remain format-compliant (well-formed bibliographic fields). Unresolved outcomes dominate (36-61%); a 100-citation audit indicates that a substantial fraction of Unresolved cases are fabricated. Results motivate post-hoc citation verification before LLM outputs enter SE literature reviews or tooling pipelines.