As a second-year PhD student, I recently submitted my first paper to the prestigious NeurIPS conference. The excitement of receiving an initial score of 4454 with a confidence score of 4334 was quickly replaced by anxiety as I entered the rebuttal phase. With no one in my lab having experience with NeurIPS, I felt overwhelmed and uncertain about how to proceed.
Luckily, I stumbled upon some helpful resources, including a Medium post on how to write rebuttals, the NeurIPS FAQs, and previous ICLR open reviews. However, I still had some questions and concerns that I hoped the machine learning community could help with.
One-Page PDF: What to Include?
NeurIPS allows authors to submit a one-page PDF with additional results. My question was: what’s the best approach for this? Should I include only figures and tables, with descriptions in the external comments, or add more text to the PDF? I’m still unsure about the optimal strategy, but I’d love to hear from others who have gone through this process.
Dealing with Requests for More Experiments
Two reviewers asked about the performance of my reinforcement learning method on long-horizon tasks. While I acknowledge that my method might not excel in these scenarios, it’s not the primary focus of my paper. The rebuttal phase is short, and building a new environment, figuring out hyperparameters, and conducting additional ablation studies would be a significant undertaking. Should I politely decline these requests, or is there a better way to address them?
These are just a few of the challenges I’m facing during the rebuttal phase. If you’ve been through this process before, I’d appreciate any guidance or advice you can offer.