Ankit Sharma, Nachiket Tapas, Jyotiprakash Patra

Improving LLM Reliability through Hybrid Abstention and Adaptive Detection

arXiv:2602.15391v1 Announce Type: new Abstract: Large Language Models (LLMs) deployed in production environments face a fundamental safety-utility trade-off either a strict filtering mechanisms prevent harmful …