When Do Hallucinations Arise? A Graph Perspective on the Evolution of Path Reuse and Path Compression

Computer Science > Artificial Intelligence [Submitted on 4 Apr 2026] When Do Hallucinations Arise? A Graph Perspective on the Evolution of Path Reuse and Path Compression Xinnan Dai, Kai Yang, Cheng Luo, Shenglai Zeng, Kai Guo, Jiliang Tang Reasoning hallucinations in large language models (LLMs) often appear as fluent yet unsupported conclusions that violate either the given context or underlying factual knowledge. Although such failures are widely observed, the mechanisms by which decoder-only Transformers produce them remain poorly understood. We model next-token prediction as a graph search process over an underlying graph, where entities correspond to nodes and learned transitions form edges. From this perspective, contextual reasoning is a constrained search over a sampled subgraph (intrinsic reasoning), while context-free queries rely on memorized structures in the underlying graph (extrinsic reasoning). We show that reasoning hallucinations arise from two fundamental mechanisms: \textbf{Path Reuse}, where memorized knowledge overrides contextual constraints during early training, and \textbf{Path Compression}, where frequently traversed multi-step paths collapse into shortcut edges in later training. Together, these mechanisms provide a unified explanation for reasoning hallucinations in LLMs and connected to well-known behaviors observed in downstream applications. Subjects: Artificial Intelligence (cs.AI) Cite as: arXiv:2604.03557 [cs.AI]   (or arXiv:2604.03557v1 [cs.AI] for this version)   https://doi.org/10.48550/arXiv.2604.03557 Focus to learn more Submission history From: Xinnan Dai [view email] [v1] Sat, 4 Apr 2026 02:51:46 UTC (4,689 KB) Access Paper: HTML (experimental) view license Current browse context: cs.AI < prev   |   next > new | recent | 2026-04 Change to browse by: cs References & Citations NASA ADS Google Scholar Semantic Scholar Export BibTeX Citation Bookmark Bibliographic Tools Bibliographic and Citation Tools Bibliographic Explorer Toggle Bibliographic Explorer (What is the Explorer?) Connected Papers Toggle Connected Papers (What is Connected Papers?) Litmaps Toggle Litmaps (What is Litmaps?) scite.ai Toggle scite Smart Citations (What are Smart Citations?) Code, Data, Media Demos Related Papers About arXivLabs Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?)