עד כמה יציבים LLM בסירוב? מדידת הבלבול בגבולות הבטיחות

תקציר מקורי באנגליתarXiv:2512.01037v3 Announce Type: replace Abstract: As safety alignment becomes standard in large language models, refusal behavior has become an important part of model reliability. However, models may still reject benign prompts, especially when the wording resembles risky content. Existing evaluations usually report global scores, such as false rejection rate or compliance rate. These scores are useful, but they treat each prompt independently. As a result, they miss local inconsistency, where a model accepts one phrasing of an intent but rejects a close paraphrase. This makes it difficult to understand whether refusals are only frequent or also semantically unstable. We address this gap by introducing Semantic Confusion, a failure mode that captures contradictory refusal decisions acro