# Post-Training BatchNorm Recalibration

arXiv: https://arxiv.org/abs/2010.05625 · HTML: https://mlsanity.com/paper/2010.05625/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 12 Oct 2020.

## Built on by (most cited first, 4 of 4)

- [Rethinking "Batch" in BatchNorm](https://mlsanity.com/paper/2105.07576/) (2021)
- [Post-Training Sparsity-Aware Quantization](https://mlsanity.com/paper/2105.11010/) (2021)
- [DeepWeightFlow: Re-Basined Flow Matching for Generating Neural Network Weights](https://mlsanity.com/paper/2601.05052/) (2026)
- [Efficient INT8 Single-Image Super-Resolution via Deployment-Aware Quantization and Teacher-Guided Training](https://mlsanity.com/paper/2604.20291/) (2026)

## Abstract (arXiv)

We revisit non-blocking simultaneous multithreading (NB-SMT) introduced previously by Shomron and Weiser (2020). NB-SMT trades accuracy for performance by occasionally "squeezing" more than one thread into a shared multiply-and-accumulate (MAC) unit. However, the method of accommodating more than one thread in a shared MAC unit may contribute noise to the computations, thereby changing the internal statistics of the model. We show that substantial model performance can be recouped by post-training recalibration of the batch normalization layers' running mean and running variance statistics, given the presence of NB-SMT.