# What Physics do Data-Driven MoCap-to-Radar Models Learn?

arXiv: https://arxiv.org/abs/2605.00018 · HTML: https://mlsanity.com/paper/2605.00018/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 8 of its 23 references that are in the map.
- First posted on arXiv 19 Apr 2026.

## Builds on (most cited first, 8 of 8)

- [The Mythos of Model Interpretability](https://mlsanity.com/paper/1606.03490/) (2016)
- [Mechanistic Interpretability for AI Safety -- A Review](https://mlsanity.com/paper/2404.14082/) (2024)
- [Interpretable Machine Learning -- A Brief History, State-of-the-Art and Challenges](https://mlsanity.com/paper/2010.09337/) (2020)
- [Discovering physical concepts with neural networks](https://mlsanity.com/paper/1807.10300/) (2018)
- [ComPhy: Compositional Physical Reasoning of Objects and Events from Videos](https://mlsanity.com/paper/2205.01089/) (2022)
- [Filtered-CoPhy: Unsupervised Learning of Counterfactual Physics in Pixel Space](https://mlsanity.com/paper/2202.00368/) (2022)
- [One Size Does Not Fit All: Multi-Scale, Cascaded RNNs for Radar Classification](https://mlsanity.com/paper/1909.03082/) (2019)
- [MoCap2Radar: A Spatiotemporal Transformer for Synthesizing Micro-Doppler Radar Signatures from Motion Capture](https://mlsanity.com/paper/2511.11462/) (2025)

## Built on by (most cited first, 1 of 1)

- [LAWFUL: Law-Aligned Witness for Faithful Use of Latents](https://mlsanity.com/paper/2607.28672/) (2026)

## Abstract (arXiv)

Data-driven MoCap-to-radar models generate plausible micro-Doppler spectrograms, but do they actually learn the underlying physics? We introduce a physics-based interpretability framework to answer this question via two proposed complementary metrics: one measures alignment between model predictions and the physics-derived Doppler frequency, while the other tests whether predictions preserve the velocity-frequency relationship under velocity intervention. Both metrics require only MoCap input and model predictions, without access to measured radar data. Experiments across several model architectures reveal that low reconstruction error does not guarantee physical consistency: some, but not all, models achieve low error yet perform poorly on the two physics-based metrics. Further analysis shows that temporal attention is critical for transformer-based models to learn the underlying physics.