Although reinforcement learning (RL) can effectively enhance the reasoning capabilities of vision–language models (VLMs), current methods remain heavily dependent on labor-intensive datasets that ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results