# Compositionality Unlocks Deep Interpretable Models

arXiv: https://arxiv.org/abs/2504.02667 · HTML: https://mlsanity.com/paper/2504.02667/

## Key facts

- Cited by 6 papers in the map (data through 14 Oct 2026).
- Builds on 11 of its 19 references that are in the map.
- First posted on arXiv 3 Apr 2025.

## Builds on (most cited first, 11 of 11)

- [GLU Variants Improve Transformer](https://mlsanity.com/paper/2002.05202/) (2020)
- [Why does deep and cheap learning work so well?](https://mlsanity.com/paper/1608.08225/) (2016)
- [QNLP in Practice: Running Compositional Models of Meaning on a Quantum Computer](https://mlsanity.com/paper/2102.12846/) (2021)
- [Deep Polynomial Neural Networks](https://mlsanity.com/paper/2006.13026/) (2020)
- [Towards falsifiable interpretability research](https://mlsanity.com/paper/2010.12016/) (2020)
- [Quantum Entanglement in Deep Learning Architectures](https://mlsanity.com/paper/1803.09780/) (2018)
- [Bilinear MLPs enable weight-based mechanistic interpretability](https://mlsanity.com/paper/2410.08417/) (2024)
- [What is the Relationship between Tensor Factorizations and Circuits (and How Can We Exploit it)?](https://mlsanity.com/paper/2409.07953/) (2024)
- [A technical note on bilinear layers for interpretability](https://mlsanity.com/paper/2305.03452/) (2023)
- [Compositionality as we see it, everywhere around us](https://mlsanity.com/paper/2110.05327/) (2021)
- [Learning with tree-based tensor formats](https://mlsanity.com/paper/1811.04455/) (2018)

## Built on by (most cited first, 6 of 6)

- [Towards Interpretability Without Sacrifice: Faithful Dense Layer Decomposition with Mixture of Decoders](https://mlsanity.com/paper/2505.21364/) (2025)
- [Finding Manifolds With Bilinear Autoencoders](https://mlsanity.com/paper/2510.16820/) (2025)
- [PolySAE: Modeling Feature Interactions in Sparse Autoencoders via Polynomial Decoding](https://mlsanity.com/paper/2602.01322/) (2026)
- [Bilinear autoencoders find interpretable manifolds](https://mlsanity.com/paper/2605.08891/) (2026)
- [From Mechanistic to Compositional Interpretability](https://mlsanity.com/paper/2605.08934/) (2026)
- [Benign Loss Landscapes Can Coexist with Worst-Case Hardness](https://mlsanity.com/paper/2609.13057/) (2026)

## Abstract (arXiv)

We propose chi-net, an intrinsically interpretable architecture combining the compositional multilinear structure of tensor networks with the expressivity and efficiency of deep neural networks. chi-nets retain equal accuracy compared to their baseline counterparts. Our novel, efficient diagonalisation algorithm, ODT, reveals linear low-rank structure in a multilayer SVHN model. We leverage this toward formal weight-based interpretability and model compression.