Question about the comparasion #15
Description
Hi, thanks for the work. From my aspect of view, Qwen2VL-Ins or other base model haven't trained on SAT or likely data at all. Does training directly on SAT with R1 really convincing
{{ message }}
Hi, thanks for the work. From my aspect of view, Qwen2VL-Ins or other base model haven't trained on SAT or likely data at all. Does training directly on SAT with R1 really convincing