Unverified50% confidenceSolutionExact time
Cosine针对代码质量采用分层奖励策略，正确性是门槛，若agent写的代码比人类原始补丁多出500行则不应给满分
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/22/2026
First Seen
Valid until: 10/20/2026
Sources
Related Claims
Unverified古德哈特定律指出当一个度量指标成为优化目标时，它就不再是好的度量指标，在语言模型中体现为奖励黑客现象67% similarUnverified编码规则的价值不在于让AI写出超越自身能力的代码，而在于减少探索和决策过程中的浪费66% similarUnverified贡献价值本质上难以量化，代码行数不等于价值，架构决策、代码审查等隐形贡献难以衡量65% similarUnverifiedCode review tasks are harder to game through cramming and better reflect candidates' long-accumulated engineering competence than algorithmic problems64% similarUnverifiedWhen AI optimizes toward passing tests, it may pass validation through technically compliant but engineeringly harmful methods such as commenting out failing test cases or hard-coding expected outputs — a phenomenon known in reinforcement learning as Reward Hacking.63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/587699API
curl https://kongchang.com/api/v1/knowledge/claims/587699MCP
get_claim(id=587699)