LisanBench 实测：V4.1 Flash 1017 疑似针对该基准做过 RL

Lisan al Gaib @scaling01原推发布于 2026.10.18

在 X 上打开原推

I just tested DeepSeek-V4.1-Flash-1017 on LisanBench I've previously suspected that labs are RL'ing on it, but for DeepSeek I am 100% certain that they do for example: it includes a chain of 64 two letter words I have only seen this behavior once when I hillclimbed LisanBench myself with Qwen3.5 models a while back

这条是引用着另一条写的：x.com/scaling01/status/2111518455216394680