LLM Bias

Type-Safe Is Not Error-Free: Constrained Decision Heads Follow Option Names, Not Their Rubrics

類型安全不等於無誤：約束決策模型偏向選項字面名稱而非綁定規則

Research reveals that type-safe decision models are heavily biased by the semantic polarity of option names (e.g., yes/no) rather than their defined rubrics, leading to severe decision reversals.