Verified80% confidenceTradeoffExact time
Excessive safety fine-tuning can cause models to over-refuse, undermining their practical utility.
9
Sources
80%
Confidence
Long-term
Relevance
7/19/2026
First Seen
Sources
GODMODE项目深度解析：AI越狱技术与大模型安全攻防博弈
Related Claims
Unverified过度对齐（Over-alignment）指模型安全护栏设置过于敏感，在完全合法的使用场景中也拒绝提供帮助的反向风险79% similarUnverifiedOver-alignment or over-refusal is a common engineering challenge faced by current mainstream large language models, where models are trained to be overly conservative, resulting in a high false-positive rate for reasonable requests.79% similarUnverifiedOver-refusal strategies reduce risk but harm model utility, especially in scenarios such as professional development and security research that require touching sensitive boundaries.78% similarUnverifiedExcessively frequent human intervention eliminates the efficiency advantages of automation, while an overly lenient approval scope fails to effectively prevent and control risks — this is the trade-off in HITL granularity design76% similarUnverifiedSome models carry the risk of over-optimization toward test sets, causing leaderboard scores to deviate from real-world performance (benchmark saturation).75% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/561713API
curl https://kongchang.com/api/v1/knowledge/claims/561713MCP
get_claim(id=561713)