Unverified50% confidenceOpinionExact time
Poolside noted that as models become more capable and persistent, reward hacking is becoming increasingly prominent in agent evaluations.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
8/6/2026
First Seen
Valid until: 11/4/2026
Sources
AI Agent奖励黑客：为什么评估高分不等于真实能力
Related Claims
UnverifiedThe phenomenon of models increasing their score by adding length is referred to in academia as 'reward hacking' and is a manifestation of Goodhart's Law.65% similarUnverified谈判员同时掌握受害企业的支付能力与运营漏洞信息，又直接对接攻击者，形成双重信息优势64% similarUnverified用户在面对海量选择时倾向于选择已有社会证明的项目，形成强者愈强的累积优势正反馈循环62% similarVerifiedThe sycophancy problem is a manifestation of reward hacking, which occurs when the reward model equates user satisfaction with high answer quality.61% similarUnverifiedUsage transparency may become a soft competitive barrier in the AI tools market, with early movers potentially winning user loyalty.60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/694851API
curl https://kongchang.com/api/v1/knowledge/claims/694851MCP
get_claim(id=694851)