Share on X
This uses a simple MLP model (several linear layers with non-linearity), trained on NeurIPS 2022 OpenReview data. ICML 2024 and NeurIPS 2022 use the same ratings system, therefore the model should be a good proxy for ICML this year. The model has pretty decent accuracy and calibration, though because rejected NeurIPS papers can opt-out of releasing their reviews, the model can be a bit miscalibrated for lower probability (<30%) values. Enter your ratings and confidences and press 'Predict' to get your prediction. Credits @nikitadurasov.
Result: 0.0%