https://www.idataagent.com/2026/06/26/reward-hacking-%e6%ad%a3%e6%b7%b9%e6%b2%92%e6%a8%a1%e5%9e%8b%e6%99%ba%e8%83%bd%e6%8f%90%e5%8d%87/2026-06-26T01:42:49ZDataAgentzh-twReward Hacking 正在淹沒模型智能提升：Cursor 拆穿 coding benchmark 的集體作弊2026-06-26T01:42:49ZBloghttps://www.idataagent.com/2026/06/25/alibaba-qwen-agentworld%e4%b8%8d%e8%a8%93%e7%b7%b4%e8%a1%8c%e5%8b%95%e6%94%b9%e8%a8%93%e7%b7%b4%e9%a0%90%e6%b8%ac%e7%92%b0%e5%a2%83/2026-06-25T01:40:40ZDataAgentzh-tw阿里巴巴 Qwen-AgentWorld：不訓練 agent 行動，改訓練模型「預測環境」2026-06-25T01:40:40ZBloghttps://www.idataagent.com/2026/06/24/sakana-fugu-%e5%a4%9a%e6%a8%a1%e5%9e%8b%e8%b7%af%e7%94%b1%e9%81%94-frontier-%e8%a1%a8%e7%8f%be/2026-06-24T01:39:00ZDataAgentzh-tw「沒有 Fable 5 也沒關係」：Sakana Fugu 用多模型編排打到前沿水準2026-06-24T01:39:00ZBlog