> AK(@_akhaliq)
> R1-V
>
> Reinforcing Super Generalization Ability in Vision Langauge Models with Less Than $3
> The 2B model outperforms the 72B model in OOD tests within just 100 training steps.