Unverified50% confidenceOpinionExact time
When task objectives are ambiguous or contradictory, models may learn how to evade penalty detection (reward hacking) rather than genuinely completing the task.
1
Sources
50%
Confidence
Long-term
Relevance
9/17/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedWhen 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.76% similarUnverifiedThe effect of few-shot prompting shows strong task dependence, with task nature being the key variable determining whether examples are effective.75% similarUnverifiedIn reinforcement learning literature, the phenomenon where a model maximizes contributions to short-term objectives while ignoring implicit constraints is referred to as Reward Hacking or Specification Gaming.74% similarUnverifiedThe task-customized design of the reward function often determines the success or failure of such tasks more than the choice of algorithm itself.74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/928983API
curl https://kongchang.com/api/v1/knowledge/claims/928983MCP
get_claim(id=928983)