Hidden Reliability Risks in Large Language Models: Systematic Identification of Precision-Induced Output Disagreements

arXiv:2604.19790v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly deployed under diverse numerical precision configurations, including standard floating-point formats (e.g., bfloat16 and float16) and quantized integer formats (e.g., int16 and int8), to meet efficiency...