Improving LLM Reliability through Hybrid Abstention and Adaptive Detection

Ankit Sharma, Nachiket Tapas, Jyotiprakash Patra

Large Language Models (LLMs) deployed in production environments face a fundamental safety-utility trade-off either a strict filtering mechanisms prevent harmful outputs but often…