精选70Rohan Paul模型发布/更新多源精选 ×7
GLM-5.3 发现 Cursor 严重漏洞，CyberGym 得分 84.5%
So GLM-5.3 has already found a "potentially serious vulnerability" in Cursor.
So GLM-5.3 has already found a "potentially serious vulnerability" in Cursor.
因此，GLM-5.3 已经在 Cursor 中发现了一个“潜在严重漏洞”。
GLM-5.3's CyberGym score rose to 84.5%, while ExploitBench more than doubled from 24.4% to 54.4%.
GLM-5.3 的 CyberGym 得分升至 84.5%，而 ExploitBench 的得分从 24.4% 翻倍以上至 54.4%。
Shows how much more performance a frontier-scale base model can deliver without going through another costly pretraining run.
这表明，在不进行另一次昂贵的预训练的情况下，前沿规模的基座模型能带来多大的性能提升。
“Scaling post-training is all we did for GLM-5.3,” Z .ai said in its technical announcement.
“我们为 GLM-5.3 所做的全部就是扩展后训练，”Z.ai 在其技术公告中表示。
更进一步：量化金融体系
看懂新闻只是起点——沿量化金融路径，把它变成能交付的工程能力