# OceanCBM: A Concept Bottleneck Model for Mechanistic Interpretability in Ocean Forecasting

arXiv: https://arxiv.org/abs/2605.12639 · HTML: https://mlsanity.com/paper/2605.12639/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Placed in Interpreting time-series and weather models, one of 13 directions in Mechanistic interpretability.
- Builds on 20 of its 45 references that are in the map.
- First posted on arXiv 12 May 2026.

## Placed in

- Mechanistic interpretability / [Interpreting time-series and weather models](https://mlsanity.com/mechinterp/interpreting-time-series-and-weather-models/)

## Builds on (most cited first, 20 of 20)

- [U-Net: Convolutional Networks for Biomedical Image Segmentation](https://mlsanity.com/paper/1505.04597/) (2015)
- [Concept Bottleneck Models](https://mlsanity.com/paper/2007.04612/) (2020)
- [Characterizing possible failure modes in physics-informed neural networks](https://mlsanity.com/paper/2109.01050/) (2021)
- [Interpretable Machine Learning: Fundamental Principles and 10 Grand Challenges](https://mlsanity.com/paper/2103.11251/) (2021)
- [Post-hoc Concept Bottleneck Models](https://mlsanity.com/paper/2205.15480/) (2022)
- [Label-Free Concept Bottleneck Models](https://mlsanity.com/paper/2304.06129/) (2023)
- [Integrating Scientific Knowledge with Machine Learning for Engineering and Environmental Systems](https://mlsanity.com/paper/2003.04919/) (2020)
- [Do Concept Bottleneck Models Learn as Intended?](https://mlsanity.com/paper/2105.04289/) (2021)
- [Rethinking Interpretability in the Era of Large Language Models](https://mlsanity.com/paper/2402.01761/) (2024)
- [Concept Bottleneck Model with Additional Unsupervised Concepts](https://mlsanity.com/paper/2202.01459/) (2022)
- [Machine learning with data assimilation and uncertainty quantification for dynamical systems: a review](https://mlsanity.com/paper/2303.10462/) (2023)
- [Will Artificial Intelligence supersede Earth System and Climate Models?](https://mlsanity.com/paper/2101.09126/) (2021)
- [Samudra: An AI Global Ocean Emulator for Climate](https://mlsanity.com/paper/2412.03795/) (2024)
- [Bridging observation, theory and numerical simulation of the ocean using Machine Learning](https://mlsanity.com/paper/2104.12506/) (2021)
- [SamudrACE: Fast and Accurate Coupled Climate Modeling with 3D Ocean and Atmosphere Emulators](https://mlsanity.com/paper/2509.12490/) (2025)
- [Data-driven Global Ocean Modeling for Seasonal to Decadal Prediction](https://mlsanity.com/paper/2405.15412/) (2024)
- [Southern Ocean Dynamics Under Climate Change: New Knowledge Through Physics-Guided Machine Learning](https://mlsanity.com/paper/2310.13916/) (2023)
- [Interpretability for Time Series Transformers using A Concept Bottleneck Framework](https://mlsanity.com/paper/2410.06070/) (2024)
- [Process-Guided Concept Bottleneck Model](https://mlsanity.com/paper/2601.10562/) (2026)
- [Data Driven Deep Learning for Correcting Global Climate Model Projections of SST and DSL in the Bay of Bengal](https://mlsanity.com/paper/2504.20620/) (2025)

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

- [Sparse probes and murky physics: a case study of interpretability challenges in a foundation model for continuum dynamics](https://mlsanity.com/paper/2606.11657/) (2026)

## Abstract (arXiv)

Extreme ocean phenomena are challenging not only to predict but to diagnose, as accurate forecasts alone do not reveal the underlying physical drivers. While recent machine learning approaches achieve strong predictive skill, they remain largely opaque and provide limited guarantees of fidelity to ground-truth physics. We introduce OceanCBM, the first concept bottleneck model (CBM) for spatiotemporal prediction and mechanistic interrogation of ocean dynamics. OceanCBM uses mixed supervision to predict mixed layer heat content, a key precursor of marine heatwaves, while routing information through an intermediate layer of prescribed concepts derived from geophysical fluid dynamics and a 'free' concept. This design imposes soft physical structure without over-constraining the model, and the free concept both regularizes concept predictions and captures residual physical processes. Across ensemble initializations, we show that mixed supervision yields consistent mechanistic representations, whereas prediction-only and prescription-only baselines learn highly variable latent structures despite similar predictive performance. OceanCBM achieves interpretable, physically grounded representations without sacrificing skill, explicitly characterizing the interpretability-performance trade-off.