Working paper2025–2026Archive record
Epistemic Risk Surfaces in Large Language Models
Abstract
Develops a taxonomy of epistemic failure in large language models, distinguishing confident error, synthetic coherence, uncertainty compression, and context-sensitive reliability collapse.
epistemic riskLLMreliability
Status of this record
This record is filed as working paper and carries the status “Working paper”. The archive states the stage of an argument instead of presenting drafts as settled results, so a reader can judge how much weight the text is meant to carry.
Citation
Ziganshin, A. (2025–2026). Epistemic Risk Surfaces in Large Language Models. Traceremove Research. https://traceremove.dev/publications/epistemic-risk-surfaces-llms