The rebuttal is where a paper's fate often feels like it hangs in the balance, yet the process remains stubbornly opaque. For researchers who received low scores at KDD and are now wondering whether a response is worth the effort, the honest answer is that the system is not designed to reward your best arguments. It is designed to test your resilience. Reviewers rarely move their scores after discussion, and when they do, it is usually by a single point, not a transformation from reject to accept. That is not cynicism; it is the accumulated experience of countless submission cycles.
We should separate two questions that the community frequently conflates. The first is whether rebuttals ever change outcomes. They do, but only at the margins, typically when a reviewer has misunderstood a technical contribution or when a factual error can be corrected with evidence. The second question is whether rebuttals are worth the emotional and intellectual investment. That answer depends on what you are optimizing for. If you are writing a rebuttal to convert a 3 into a 5, you are likely wasting your time. If you are writing it to clarify the record for the meta-reviewer, or to sharpen your own understanding of the paper's weaknesses before the next submission, the exercise has real value. The same logic applies to the broader research ecosystem, where navigating funding gaps when an accepted paper needs to withdraw teaches us that the administrative reality of publishing often matters more than the score itself.
What makes the KDD case particularly frustrating is the asymmetry of information. Reviewers are not obligated to respond to rebuttals, and many do not. You can submit a careful, point-by-point response and hear nothing back. Then the decision arrives with boilerplate language that does not acknowledge your arguments. That silence is not a failure of your writing; it is a structural feature of a review process where discussion is optional. We have seen this pattern in other technical communities, where the Linux Foundation’s push for asymmetric multiprocessing contributors highlights how even well-organized technical bodies struggle to align incentives across diverse stakeholders. When participation is voluntary and unenforced, the quality of engagement degrades.
The practical takeaway for anyone facing a low KDD score is to submit a rebuttal only if you can identify a specific, addressable misunderstanding. Do not write a general defense of your paper's significance. Do not argue with a reviewer's taste. Point to the exact line in the paper that addresses their concern, and state plainly what they missed. Then move on. The rebuttal is not a negotiation; it is a formality that occasionally corrects errors. If you are expecting a dialogue, you will be disappointed. If you are expecting a chance to demonstrate how small AI models closed the gap on a test built for humans by learning from repeated failures, you will find the analogy apt. The system rewards persistence across submissions, not persuasion within a single cycle. Watch how the meta-reviewers treat rebuttals in the next round of decisions. That will tell you more about the future of KDD than any single score change.