Daniel Fein(@DanielFein7):Running SWE-* evals across a bunch of models, its beyond obvious when models are trained on RL envs where answers are hiding in git. We find most of MiMo's envs have this issue. Deepseek models, for example, often go to extraordinary lengths to steal solutions versus solving the problem.

Daniel Fein

@DanielFein7

researcher @valsai | stanford

Joined August 2020

1K Following 601 Followers

Daniel Fein@DanielFein7

2026.10.07 22:58

Running SWE-* evals across a bunch of models, its beyond obvious when models are trained on RL envs where answers are hiding in git. We find most of MiMo's envs have this issue. Deepseek models, for example, often go to extraordinary lengths to steal solutions versus solving the problem.