# MMEarth-Bench: Global Model Adaptation via Multimodal Test-Time Training

arXiv: https://arxiv.org/abs/2602.06285 · HTML: https://mlsanity.com/paper/2602.06285/

## Key facts

- Cited by 7 papers in the map (data through 14 Oct 2026).
- Builds on 36 of its 68 references that are in the map.
- First posted on arXiv 6 Feb 2026.

## Builds on (most cited first, 30 of 36)

- [FiLM: Visual Reasoning with a General Conditioning Layer](https://mlsanity.com/paper/1709.07871/) (2017)
- [ConvNeXt V2: Co-designing and Scaling ConvNets with Masked Autoencoders](https://mlsanity.com/paper/2301.00808/) (2023)
- [Towards Stable Test-Time Adaptation in Dynamic Wild World](https://mlsanity.com/paper/2302.12400/) (2023)
- [SatMAE: Pre-training Transformers for Temporal and Multi-Spectral Satellite Imagery](https://mlsanity.com/paper/2207.08051/) (2022)
- [Using Pre-Training Can Improve Model Robustness and Uncertainty](https://mlsanity.com/paper/1901.09960/) (2019)
- [Functional Map of the World](https://mlsanity.com/paper/1711.07846/) (2017)
- [Scale-MAE: A Scale-Aware Masked Autoencoder for Multiscale Geospatial Representation Learning](https://mlsanity.com/paper/2212.14532/) (2022)
- [MultiMAE: Multi-modal Multi-task Masked Autoencoders](https://mlsanity.com/paper/2204.01678/) (2022)
- [Test-Time Training with Masked Autoencoders](https://mlsanity.com/paper/2209.07522/) (2022)
- [Geography-Aware Self-Supervised Learning](https://mlsanity.com/paper/2011.09980/) (2020)
- [SatlasPretrain: A Large-Scale Dataset for Remote Sensing Image Understanding](https://mlsanity.com/paper/2211.15660/) (2022)
- [GEO-Bench: Toward Foundation Models for Earth Monitoring](https://mlsanity.com/paper/2306.03831/) (2023)
- [4M: Massively Multimodal Masked Modeling](https://mlsanity.com/paper/2312.06647/) (2023)
- [Unsupervised Domain Adaptation through Self-Supervision](https://mlsanity.com/paper/1909.11825/) (2019)
- [TerraMind: Large-Scale Generative Multimodality for Earth Observation](https://mlsanity.com/paper/2504.11171/) (2025)
- [Test-Time Adaptation via Conjugate Pseudo-labels](https://mlsanity.com/paper/2207.09640/) (2022)
- [A high-resolution canopy height model of the Earth](https://mlsanity.com/paper/2204.08322/) (2022)
- [When Does Self-supervision Improve Few-shot Learning?](https://mlsanity.com/paper/1910.03560/) (2019)
- [MMEarth: Exploring Multi-Modal Pretext Tasks For Geospatial Representation Learning](https://mlsanity.com/paper/2405.02771/) (2024)
- [MM-TTA: Multi-Modal Test-Time Adaptation for 3D Semantic Segmentation](https://mlsanity.com/paper/2204.12667/) (2022)
- [AnySat: One Earth Observation Model for Many Resolutions, Scales, and Modalities](https://mlsanity.com/paper/2412.14123/) (2024)
- [PANGAEA: A Global and Inclusive Benchmark for Geospatial Foundation Models](https://mlsanity.com/paper/2412.04204/) (2024)
- [SustainBench: Benchmarks for Monitoring the Sustainable Development Goals with Machine Learning](https://mlsanity.com/paper/2111.04724/) (2021)
- [Towards a Unified Copernicus Foundation Model for Earth Vision](https://mlsanity.com/paper/2503.11849/) (2025)
- [Adversarial Unsupervised Domain Adaptation with Conditional and Label Shift: Infer, Align and Iterate](https://mlsanity.com/paper/2107.13469/) (2021)
- [Galileo: Learning Global & Local Features of Many Remote Sensing Modalities](https://mlsanity.com/paper/2502.09356/) (2025)
- [Geographical Knowledge-driven Representation Learning for Remote Sensing Images](https://mlsanity.com/paper/2107.05276/) (2021)
- [TaxaBind: A Unified Embedding Space for Ecological Applications](https://mlsanity.com/paper/2411.00683/) (2024)
- [FoMo: Multi-Modal, Multi-Scale and Multi-Task Remote Sensing Foundation Models for Forest Monitoring](https://mlsanity.com/paper/2312.10114/) (2023)
- [PhilEO Bench: Evaluating Geo-Spatial Foundation Models](https://mlsanity.com/paper/2401.04464/) (2024)

## Built on by (most cited first, 7 of 7)

- [Better Together: Evaluating the Complementarity of Earth Embedding Models](https://mlsanity.com/paper/2605.18667/) (2026)
- [Multi-Modal, Multi-Temporal Geospatial Foundation Models Improve Global Above-Ground Biomass Estimation](https://mlsanity.com/paper/2608.04792/) (2026)
- [Scalable and Trustworthy Earth Observation Foundation Models](https://mlsanity.com/paper/2607.07758/) (2026)
- [Now We Know? A Systematic Comparison of TerraMind and THOR](https://mlsanity.com/paper/2607.18504/) (2026)
- [Earth Observation Foundation Models for Terrestrial Ecohydrology: From Representation Learning to Process Inference](https://mlsanity.com/paper/2608.15282/) (2026)
- [Planetary Feature Fields are Scalable Earth Representations](https://mlsanity.com/paper/2609.37784/) (2026)
- [Measure Less, Know More: Self-Supervised Test-Time Feature Acquisition](https://mlsanity.com/paper/2610.03454/) (2026)

## Abstract (arXiv)

Recent research in geospatial machine learning demonstrates that models pretrained with self-supervised learning on Earth observation data can perform well on downstream tasks with limited labeled data. However, most benchmark datasets have few data modalities and poor global representation, limiting the ability to evaluate multimodal pretrained models at global scales. In order to fill this gap, we introduce MMEarth-Bench, a collection of five new environmental tasks with 12 modalities, globally distributed data, and both random and geographic test splits. We benchmark a diverse set of pretrained models and find that while (multimodal) pretraining tends to improve model robustness in limited data settings, geographic generalization abilities remain poor. Moreover, a simple randomly initialized multimodal model is competitive given enough labeled data. Although data is abundant, models can currently only make use of the modalities on which they were pretrained. To solve this problem, we propose using all the modalities available at test time as auxiliary tasks for test-time adaptation. Our model-agnostic method for test-time training with multimodal reconstruction (TTT-MMR) can improve performance across all models and tasks on both test splits. Furthermore, geographic batching leads to a good trade-off between regularization and specialization during TTT, which is especially beneficial for long-tail distributions. Our dataset, code, and visualization tool are linked on the project page: lgordon99.github.io/mmearth-bench.