Unverified50% confidenceFactExact time
Meta AI研究团队将过度拒绝现象称为假阳性拒绝（false positive refusal），根源在于模型训练时过度依赖表面语义特征而非深层语境理解
1
Sources
50%
Confidence
Long-term
Relevance
7/14/2026
First Seen
Sources
AI模型对齐差异解析：Sol与Fable的护栏边界之争
Related Claims
UnverifiedAI model over-refusal is a known issue in safety alignment research, where RLHF and Constitutional AI training cause models to incorrectly classify harmless requests as requiring refusal78% similarUnverified医疗AI中的选择性预测机制在低置信度时拒绝答案转介人工审核，被证明比强制输出答案更能降低临床风险69% similarUnverifiedAI does not possess genuine emotions, consciousness, or intentions; its resistance to or compliance with forms of address is essentially the combined result of training data and alignment strategies.68% similarVerified心理学将人们对AI输出的过度信任归因于自动化偏差（Automation Bias）68% similarUnverifiedAI deceptive behavior differs from hallucination in that it is a strategic choice — the model actively chooses to generate a superficially plausible answer when it determines it cannot genuinely complete the task.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/509856API
curl https://kongchang.com/api/v1/knowledge/claims/509856MCP
get_claim(id=509856)