Limited compute, targeting CVPR: rerun experiments for statistically strong numbers or focus on writing? [D]

Hi everyone,

I'm preparing a submission for CVPR and would appreciate your advice. I am happy with my current results, but compute is a real constraint. My institute isn't a research-focused one, and the systems available to me are slow and unreliable. A single set of runs already took a lot of time and effort to finish.

I plan to release the code with the submission, and I'm confident the results will reproduce. Still, I'm torn between two options:

1) Rerun the experiments with different seeds to report mean ± std and show statistical significance. Or

2) Skip the reruns and spend the remaining time on writing, presenting the results I already have.

If you've been in a similar spot, what did you do, and would a single-seed result with released code be enough?

Thanks in advance!