Score your paper
Upload the anonymized PDF of your submission. PaperScorer reads its pages and emails you the model's accept or reject decision, together with a calibrated confidence in this prediction.
Your paper is in the queue
We will email the decision and its confidence to .
If the worker finds a problem with the pages or the anonymization, that email will say what to fix instead.
Submission id
Good to know
- The result is the model's accept or reject decision plus a calibrated confidence in this prediction, not a review.
- Before scoring, the worker re-checks the pages and the anonymization. If either fails, you get an email explaining what to fix.
- Only the first 15 pages are read, and the PDF can be at most 20 MB.
- Each email can submit up to 5 papers per day (UTC).
Examples of good papers
Below are well-known, good papers, anonymized and scored using PaperScorer-Vision. The line shows the model's calibrated acceptance probability, and papers above 50% are predicted accepts. Two of these papers were actually rejected. The first is an early version of A Decade's Battle on Dataset Bias (ECCV 2024), which PaperScorer-Vision correctly predicts as a reject. The second is Mamba, rejected at ICLR 2024 but scored at 84%; many would argue it deserved acceptance. We also include the newer, accepted ICLR version of A Decade's Battle on Dataset Bias, which PaperScorer-Vision correctly predicts as an accept. These papers show that PaperScorer agrees with, and sometimes defies, previous decisions, based on its estimation of quality. Click a paper to see its pages.