Lucia Gordon
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Lucia Gordon has a total of 22 co-authors. Here are the top co-authors by number of shared publications:
- Nico Lang
University of Copenhagen
3 shared publications - Andrew B. Davies
Department of Organismic and Evolutionary Biology Harvard University Cambridge Massachusetts USA
2 shared publications - Andrew Davies
Harvard University
2 shared publications Catherine Ressijac
Harvard University
2 shared publications- Catherine Ressijac
Harvard University
2 shared publications - Christian Igel
Department of Computer Science, University of Copenhagen, Copenhagen, Denmark
2 shared publications
Latest Publications
- 10 September 2026
- Published by Springer Nature
- p. 334-353
MMEarth-Bench: Global Model Adaptation via Multimodal Test-Time Training
- 10 September 2026
- Published by Springer Nature
- p. 334-353
MMEarth-Bench: Global Model Adaptation via Multimodal Test-Time Training
- Open Access
- 14 March 2026
- Published by Copernicus GmbH
MMEarth-Bench: Global Environmental Tasks for Multimodal Geospatial Models
Recent research in geospatial machine learning has demonstrated that models pretrained with self-supervised learning on Earth observation data can perform well on downstream tasks with limited training data. However, most of the existing geospatial benchmark datasets have few data modalities and poor global representation, limiting the ability to evaluate multimodal pretrained models at global scales. To fill this gap, we introduce MMEarth-Bench, a collection of five new multimodal downstream tasks with 12 input modalities, globally distributed data, and both in- and out-of-distribution test splits. We benchmark a diverse set of pretrained models on MMEarth-Bench and find that multimodal models generally perform best. While pretraining tends to improve model robustness in limited data settings, geographic generalization abilities remain poor and using multimodal inputs at test time can sometimes lead to geographic overfitting. In order to facilitate model adaptation to new downstream tasks and geographic domains, we propose a model-agnostic method for test-time training with multimodal reconstruction (TTT-MMR) that uses all the modalities available at test time, regardless of whether the pretrained model accepts them as input. We show that TTT-MMR improves model performance on both random and geographic test splits, and that geographic batching (TTT-MMR-Geo) leads to a good trade-off between regularization and specialization during TTT. Our dataset, code, and visualization tool are linked from the project page at https://lgordon99.github.io/mmearth-bench.
- 14 March 2026
- Published by Copernicus GmbH
Open AccessMMEarth-Bench: Global Environmental Tasks for Multimodal Geospatial Models
Recent research in geospatial machine learning has demonstrated that models pretrained with self-supervised learning on Earth observation data can perform well on downstream tasks with limited training data. However, most of the existing geospatial benchmark datasets have few data modalities and poor global representation, limiting the ability to evaluate multimodal pretrained models at global scales. To fill this gap, we introduce MMEarth-Bench, a collection of five new multimodal downstream tasks with 12 input modalities, globally distributed data, and both in- and out-of-distribution test splits. We benchmark a diverse set of pretrained models on MMEarth-Bench and find that multimodal models generally perform best. While pretraining tends to improve model robustness in limited data settings, geographic generalization abilities remain poor and using multimodal inputs at test time can sometimes lead to geographic overfitting. In order to facilitate model adaptation to new downstream tasks and geographic domains, we propose a model-agnostic method for test-time training with multimodal reconstruction (TTT-MMR) that uses all the modalities available at test time, regardless of whether the pretrained model accepts them as input. We show that TTT-MMR improves model performance on both random and geographic test splits, and that geographic batching (TTT-MMR-Geo) leads to a good trade-off between regularization and specialization during TTT. Our dataset, code, and visualization tool are linked from the project page at https://lgordon99.github.io/mmearth-bench.
- Open Access
MMEarth-Bench: Global Model Adaptation via Multimodal Test-Time Training
Recent research in geospatial machine learning has demonstrated that models pretrained with self-supervised learning on Earth observation data can perform well on downstream tasks with limited training data. However, most of the existing geospatial benchmark datasets have few data modalities and poor global representation, limiting the ability to evaluate multimodal pretrained models at global scales. To fill this gap, we introduce MMEarth-Bench, a collection of five new multimodal environmental tasks with 12 modalities, globally distributed data, and both in- and out-of-distribution 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. In order to facilitate model adaptation to new downstream tasks and geographic domains, we propose a model-agnostic method for test-time training with multimodal reconstruction (TTT-MMR) that uses all the modalities available at test time as auxiliary tasks, regardless of whether a pretrained model accepts them as input. Our method improves model performance on both the random and geographic test splits, and geographic batching leads to a good trade-off between regularization and specialization during TTT. Our dataset, code, and visualization tool are linked from the project page at lgordon99.github.io/mmearth-bench.
Open AccessMMEarth-Bench: Global Model Adaptation via Multimodal Test-Time Training
Recent research in geospatial machine learning has demonstrated that models pretrained with self-supervised learning on Earth observation data can perform well on downstream tasks with limited training data. However, most of the existing geospatial benchmark datasets have few data modalities and poor global representation, limiting the ability to evaluate multimodal pretrained models at global scales. To fill this gap, we introduce MMEarth-Bench, a collection of five new multimodal environmental tasks with 12 modalities, globally distributed data, and both in- and out-of-distribution 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. In order to facilitate model adaptation to new downstream tasks and geographic domains, we propose a model-agnostic method for test-time training with multimodal reconstruction (TTT-MMR) that uses all the modalities available at test time as auxiliary tasks, regardless of whether a pretrained model accepts them as input. Our method improves model performance on both the random and geographic test splits, and geographic batching leads to a good trade-off between regularization and specialization during TTT. Our dataset, code, and visualization tool are linked from the project page at lgordon99.github.io/mmearth-bench.
- 12 May 2025
- Published by Springer Nature
- p. 208-216
Multimodal Fusion Strategies for Mapping Biophysical Landscape Features
- 12 May 2025
- Published by Springer Nature
- p. 208-216
Multimodal Fusion Strategies for Mapping Biophysical Landscape Features
- Open Access
- 16 October 2024
- Published by IOS Press
- p. 3284-3291
Combining Diverse Information for Coordinated Action: Stochastic Bandit Algorithms for Heterogeneous Agents
Stochastic multi-agent multi-armed bandits typically assume that the rewards from each arm follow a fixed distribution, regardless of which agent pulls the arm. However, in many real-world settings, rewards can depend on the sensitivity of each agent to their environment. In medical screening, disease detection rates can vary by test type; in preference matching, rewards can depend on user preferences; and in environmental sensing, observation quality can vary across sensors. Since past work does not specify how to allocate agents of heterogeneous but known sensitivity of these types in a stochastic bandit setting, we introduce a UCB-style algorithm, Min-Width, which aggregates information from diverse agents. In doing so, we address the joint challenges of (i) aggregating the rewards, which follow different distributions for each agent-arm pair, and (ii) coordinating the assignments of agents to arms. Min-Width facilitates efficient collaboration among heterogeneous agents, exploiting the known structure in the agents’ reward functions to weight their rewards accordingly. We analyze the regret of Min-Width and conduct pseudo-synthetic and fully synthetic experiments to study the performance of different levels of information sharing. Our results confirm that the gains to modeling agent heterogeneity tend to be greater when the sensitivities are more varied across agents, while combining more information does not always improve performance.
- 16 October 2024
- Published by IOS Press
- p. 3284-3291
Open AccessCombining Diverse Information for Coordinated Action: Stochastic Bandit Algorithms for Heterogeneous Agents
Stochastic multi-agent multi-armed bandits typically assume that the rewards from each arm follow a fixed distribution, regardless of which agent pulls the arm. However, in many real-world settings, rewards can depend on the sensitivity of each agent to their environment. In medical screening, disease detection rates can vary by test type; in preference matching, rewards can depend on user preferences; and in environmental sensing, observation quality can vary across sensors. Since past work does not specify how to allocate agents of heterogeneous but known sensitivity of these types in a stochastic bandit setting, we introduce a UCB-style algorithm, Min-Width, which aggregates information from diverse agents. In doing so, we address the joint challenges of (i) aggregating the rewards, which follow different distributions for each agent-arm pair, and (ii) coordinating the assignments of agents to arms. Min-Width facilitates efficient collaboration among heterogeneous agents, exploiting the known structure in the agents’ reward functions to weight their rewards accordingly. We analyze the regret of Min-Width and conduct pseudo-synthetic and fully synthetic experiments to study the performance of different levels of information sharing. Our results confirm that the gains to modeling agent heterogeneity tend to be greater when the sensitivities are more varied across agents, while combining more information does not always improve performance.
- Open Access
Multimodal Fusion Strategies for Mapping Biophysical Landscape Features
Multimodal aerial data are used to monitor natural systems, and machine learning can significantly accelerate the classification of landscape features within such imagery to benefit ecology and conservation. It remains under-explored, however, how these multiple modalities ought to be fused in a deep learning model. As a step towards filling this gap, we study three strategies (Early fusion, Late fusion, and Mixture of Experts) for fusing thermal, RGB, and LiDAR imagery using a dataset of spatially-aligned orthomosaics in these three modalities. In particular, we aim to map three ecologically-relevant biophysical landscape features in African savanna ecosystems: rhino middens, termite mounds, and water. The three fusion strategies differ in whether the modalities are fused early or late, and if late, whether the model learns fixed weights per modality for each class or generates weights for each class adaptively, based on the input. Overall, the three methods have similar macro-averaged performance with Late fusion achieving an AUC of 0.698, but their per-class performance varies strongly, with Early fusion achieving the best recall for middens and water and Mixture of Experts achieving the best recall for mounds.
Open AccessMultimodal Fusion Strategies for Mapping Biophysical Landscape Features
Multimodal aerial data are used to monitor natural systems, and machine learning can significantly accelerate the classification of landscape features within such imagery to benefit ecology and conservation. It remains under-explored, however, how these multiple modalities ought to be fused in a deep learning model. As a step towards filling this gap, we study three strategies (Early fusion, Late fusion, and Mixture of Experts) for fusing thermal, RGB, and LiDAR imagery using a dataset of spatially-aligned orthomosaics in these three modalities. In particular, we aim to map three ecologically-relevant biophysical landscape features in African savanna ecosystems: rhino middens, termite mounds, and water. The three fusion strategies differ in whether the modalities are fused early or late, and if late, whether the model learns fixed weights per modality for each class or generates weights for each class adaptively, based on the input. Overall, the three methods have similar macro-averaged performance with Late fusion achieving an AUC of 0.698, but their per-class performance varies strongly, with Early fusion achieving the best recall for middens and water and Mixture of Experts achieving the best recall for mounds.
- Open Access
Find Rhinos without Finding Rhinos: Active Learning with Multimodal Imagery of South African Rhino Habitats
Much of Earth's charismatic megafauna is endangered by human activities, particularly the rhino, which is at risk of extinction due to the poaching crisis in Africa. Monitoring rhinos' movement is crucial to their protection but has unfortunately proven difficult because rhinos are elusive. Therefore, instead of tracking rhinos, we propose the novel approach of mapping communal defecation sites, called middens, which give information about rhinos' spatial behavior valuable to anti-poaching, management, and reintroduction efforts. This paper provides the first-ever mapping of rhino midden locations by building classifiers to detect them using remotely sensed thermal, RGB, and LiDAR imagery in passive and active learning settings. As existing active learning methods perform poorly due to the extreme class imbalance in our dataset, we design MultimodAL, an active learning system employing a ranking technique and multimodality to achieve competitive performance with passive learning models with 94% fewer labels. Our methods could therefore save over 76 hours in labeling time when used on a similarly-sized dataset. Unexpectedly, our midden map reveals that rhino middens are not randomly distributed throughout the landscape; rather, they are clustered. Consequently, rangers should be targeted at areas with high midden densities to strengthen anti-poaching efforts, in line with UN Target 15.7.
Open AccessFind Rhinos without Finding Rhinos: Active Learning with Multimodal Imagery of South African Rhino Habitats
Much of Earth's charismatic megafauna is endangered by human activities, particularly the rhino, which is at risk of extinction due to the poaching crisis in Africa. Monitoring rhinos' movement is crucial to their protection but has unfortunately proven difficult because rhinos are elusive. Therefore, instead of tracking rhinos, we propose the novel approach of mapping communal defecation sites, called middens, which give information about rhinos' spatial behavior valuable to anti-poaching, management, and reintroduction efforts. This paper provides the first-ever mapping of rhino midden locations by building classifiers to detect them using remotely sensed thermal, RGB, and LiDAR imagery in passive and active learning settings. As existing active learning methods perform poorly due to the extreme class imbalance in our dataset, we design MultimodAL, an active learning system employing a ranking technique and multimodality to achieve competitive performance with passive learning models with 94% fewer labels. Our methods could therefore save over 76 hours in labeling time when used on a similarly-sized dataset. Unexpectedly, our midden map reveals that rhino middens are not randomly distributed throughout the landscape; rather, they are clustered. Consequently, rangers should be targeted at areas with high midden densities to strengthen anti-poaching efforts, in line with UN Target 15.7.
- Open Access
Combining Diverse Information for Coordinated Action: Stochastic Bandit Algorithms for Heterogeneous Agents
Stochastic multi-agent multi-armed bandits typically assume that the rewards from each arm follow a fixed distribution, regardless of which agent pulls the arm. However, in many real-world settings, rewards can depend on the sensitivity of each agent to their environment. In medical screening, disease detection rates can vary by test type; in preference matching, rewards can depend on user preferences; and in environmental sensing, observation quality can vary across sensors. Since past work does not specify how to allocate agents of heterogeneous but known sensitivity of these types in a stochastic bandit setting, we introduce a UCB-style algorithm, Min-Width, which aggregates information from diverse agents. In doing so, we address the joint challenges of (i) aggregating the rewards, which follow different distributions for each agent-arm pair, and (ii) coordinating the assignments of agents to arms. Min-Width facilitates efficient collaboration among heterogeneous agents, exploiting the known structure in the agents' reward functions to weight their rewards accordingly. We analyze the regret of Min-Width and conduct pseudo-synthetic and fully synthetic experiments to study the performance of different levels of information sharing. Our results confirm that the gains to modeling agent heterogeneity tend to be greater when the sensitivities are more varied across agents, while combining more information does not always improve performance.
Open AccessCombining Diverse Information for Coordinated Action: Stochastic Bandit Algorithms for Heterogeneous Agents
Stochastic multi-agent multi-armed bandits typically assume that the rewards from each arm follow a fixed distribution, regardless of which agent pulls the arm. However, in many real-world settings, rewards can depend on the sensitivity of each agent to their environment. In medical screening, disease detection rates can vary by test type; in preference matching, rewards can depend on user preferences; and in environmental sensing, observation quality can vary across sensors. Since past work does not specify how to allocate agents of heterogeneous but known sensitivity of these types in a stochastic bandit setting, we introduce a UCB-style algorithm, Min-Width, which aggregates information from diverse agents. In doing so, we address the joint challenges of (i) aggregating the rewards, which follow different distributions for each agent-arm pair, and (ii) coordinating the assignments of agents to arms. Min-Width facilitates efficient collaboration among heterogeneous agents, exploiting the known structure in the agents' reward functions to weight their rewards accordingly. We analyze the regret of Min-Width and conduct pseudo-synthetic and fully synthetic experiments to study the performance of different levels of information sharing. Our results confirm that the gains to modeling agent heterogeneity tend to be greater when the sensitivities are more varied across agents, while combining more information does not always improve performance.
- Open Access
- 1 August 2023
- Published by International Joint Conferences on Artificial Intelligence
- Vol. 6, 5977-5985
Find Rhinos without Finding Rhinos: Active Learning with Multimodal Imagery of South African Rhino Habitats
Much of Earth's charismatic megafauna is endangered by human activities, particularly the rhino, which is at risk of extinction due to the poaching crisis in Africa. Monitoring rhinos' movement is crucial to their protection but has unfortunately proven difficult because rhinos are elusive. Therefore, instead of tracking rhinos, we propose the novel approach of mapping communal defecation sites, called middens, which give information about rhinos' spatial behavior valuable to anti-poaching, management, and reintroduction efforts. This paper provides the first-ever mapping of rhino midden locations by building classifiers to detect them using remotely sensed thermal, RGB, and LiDAR imagery in passive and active learning settings. As existing active learning methods perform poorly due to the extreme class imbalance in our dataset, we design MultimodAL, an active learning system employing a ranking technique and multimodality to achieve competitive performance with passive learning models with 94% fewer labels. Our methods could therefore save over 76 hours in labeling time when used on a similarly-sized dataset. Unexpectedly, our midden map reveals that rhino middens are not randomly distributed throughout the landscape; rather, they are clustered. Consequently, rangers should be targeted at areas with high midden densities to strengthen anti-poaching efforts, in line with UN Target 15.7.
- 1 August 2023
- Published by International Joint Conferences on Artificial Intelligence
- Vol. 6, 5977-5985
Open AccessFind Rhinos without Finding Rhinos: Active Learning with Multimodal Imagery of South African Rhino Habitats
Much of Earth's charismatic megafauna is endangered by human activities, particularly the rhino, which is at risk of extinction due to the poaching crisis in Africa. Monitoring rhinos' movement is crucial to their protection but has unfortunately proven difficult because rhinos are elusive. Therefore, instead of tracking rhinos, we propose the novel approach of mapping communal defecation sites, called middens, which give information about rhinos' spatial behavior valuable to anti-poaching, management, and reintroduction efforts. This paper provides the first-ever mapping of rhino midden locations by building classifiers to detect them using remotely sensed thermal, RGB, and LiDAR imagery in passive and active learning settings. As existing active learning methods perform poorly due to the extreme class imbalance in our dataset, we design MultimodAL, an active learning system employing a ranking technique and multimodality to achieve competitive performance with passive learning models with 94% fewer labels. Our methods could therefore save over 76 hours in labeling time when used on a similarly-sized dataset. Unexpectedly, our midden map reveals that rhino middens are not randomly distributed throughout the landscape; rather, they are clustered. Consequently, rangers should be targeted at areas with high midden densities to strengthen anti-poaching efforts, in line with UN Target 15.7.
- 24 February 2021
- Published by American Physical Society (APS) in Physical Review D
- Vol. 103 (4) , 046016
Quantum gravitational onset of Starobinsky inflation in a closed universe
Recent cosmic microwave background observations favor low-energy-scale inflationary models in a closed universe. However, the onset of inflation in such models for a closed universe is known to be severely problematic. In particular, such a universe recollapses within a few Planck seconds and encounters a big crunch singularity when initial conditions are given in the Planck regime. We show that this problem of the onset of inflation in low-energy-scale inflationary models can be successfully overcome in a quantum-gravitational framework where the big bang/big crunch singularities are resolved and a nonsingular cyclic evolution exists prior to inflation. As an example, we consider a model in loop quantum cosmology and demonstrate that the successful onset of low-energy-scale inflation in a closed universe is possible for the Starobinsky inflationary model starting from a variety of initial conditions where it is impossible in the classical theory. For comparison, we also investigate the onset of inflation in the inflationary model under highly unfavorable conditions and find similar results. Our numerical investigation including the phase-space analysis shows that the preinflationary phase with quantum gravity effects is composed of nonidentical cycles of bounces and recollapses resulting in a hysteresis-like phenomenon, which plays an important role in creating suitable conditions for inflation to occur after some number of nonsingular cycles. Our analysis shows that the tension in the classical theory amounting to the unsuitability of closed Friedmann-Lemaître-Robertson-Walker universes with respect to the onset of low-energy-scale inflation can be successfully resolved in loop quantum cosmology.- 24 February 2021
- Published by American Physical Society (APS) in Physical Review D
- Vol. 103 (4) , 046016
Quantum gravitational onset of Starobinsky inflation in a closed universe
Recent cosmic microwave background observations favor low-energy-scale inflationary models in a closed universe. However, the onset of inflation in such models for a closed universe is known to be severely problematic. In particular, such a universe recollapses within a few Planck seconds and encounters a big crunch singularity when initial conditions are given in the Planck regime. We show that this problem of the onset of inflation in low-energy-scale inflationary models can be successfully overcome in a quantum-gravitational framework where the big bang/big crunch singularities are resolved and a nonsingular cyclic evolution exists prior to inflation. As an example, we consider a model in loop quantum cosmology and demonstrate that the successful onset of low-energy-scale inflation in a closed universe is possible for the Starobinsky inflationary model starting from a variety of initial conditions where it is impossible in the classical theory. For comparison, we also investigate the onset of inflation in the inflationary model under highly unfavorable conditions and find similar results. Our numerical investigation including the phase-space analysis shows that the preinflationary phase with quantum gravity effects is composed of nonidentical cycles of bounces and recollapses resulting in a hysteresis-like phenomenon, which plays an important role in creating suitable conditions for inflation to occur after some number of nonsingular cycles. Our analysis shows that the tension in the classical theory amounting to the unsuitability of closed Friedmann-Lemaître-Robertson-Walker universes with respect to the onset of low-energy-scale inflation can be successfully resolved in loop quantum cosmology.