GPUs running deep learning (DL) workloads are frequently underutilized. Collocating multiple DL training tasks on the same GPU can improve utilization but introduces two key risks: (1) out-of-memory (OOM) crashes for newly scheduled tasks, and (2) severe performance interference among co-running tasks, which can negate any throughput gains. These issues reduce system robustness, quality of service, and energy efficiency. We present CARMA, a task-level, collocation-aware resource management system for the server-scale. CARMA addresses collocation challenges via (1) fine-grained monitoring and bookkeeping of GPUs and a collocation risk analysis that filters out the high-risk GPUs; (2) task placement policies that cap GPU utilization to avoid OOMs and limit interference; (3) integration of GPU memory need estimators for DL tasks to minimize OOMs during collocation; and (4) a lightweight recovery method that relaunches jobs crashed due to OOMs. Our evaluation on a DL training workload derived from real-world traces shows that CARMA uses GPUs more efficiently by making more informed collocation decisions: for the best-performing collocation policy, CARMA increases GPU streaming multiprocessor (SM) utilization by 54%, the parallelism achieved per SM by 61%, and memory use by 62%. This results in a $\sim$35% and $\sim$15% reduction in the end-to-end execution time (makespan) and GPU energy consumption, respectively, for this workload.

-- We introduce CARMA, a system for situational grounding in human-robot group interactions. Effective collaboration in such group settings requires situational awareness based on a consistent representation of present persons and objects coupled with an episodic abstraction of events regarding actors and manipulated objects. This calls for a clear and consistent assignment of instances, ensuring that robots correctly recognize and track actors, objects, and their interactions over time. T o achieve this, CARMA uniquely identifies physical instances of such entities in the real world and organizes them into grounded triplets of actors, objects, and actions. T o validate our approach, we conducted three experiments, where multiple humans and a robot interact: collaborative pouring, handovers, and sorting. These scenarios allow the assessment of the system's capabilities as to role distinction, multi-actor awareness, and consistent instance identification. Our experiments demonstrate that the system can reliably generate accurate actor-action-object triplets, providing a structured and robust foundation for applications requiring spatiotemporal reasoning and situated decision-making in collaborative settings.

Large language models (LLMs) struggle with compositional generalisation, limiting their ability to systematically combine learned components to interpret novel inputs. While architectural modifications, fine-tuning, and data augmentation improve compositionality, they often have limited adaptability, face scalability constraints, or yield diminishing returns on real data. To address this, we propose CARMA, an intervention that enhances the stability and robustness of compositional reasoning in LLMs while preserving fine-tuned performance. CARMA employs mutual information regularisation and layer-wise stability constraints to mitigate feature fragmentation, ensuring structured representations persist across and within layers. We evaluate CARMA on inverse dictionary modelling and sentiment classification, measuring its impact on semantic consistency, performance stability, and robustness to lexical perturbations. Results show that CARMA reduces the variability introduced by fine-tuning, stabilises token representations, and improves compositional reasoning. While its effectiveness varies across architectures, CARMA's key strength lies in reinforcing learned structures rather than introducing new capabilities, making it a scalable auxiliary method. These findings suggest that integrating CARMA with fine-tuning can improve compositional generalisation while maintaining task-specific performance in LLMs.

CARMA is an advisory system for rangeland grasshopper infestations that demonstrates how AI technology can deliver expert advice to compensate for cutbacks in public services. CARMA uses two knowledge sources for the key task of predicting forage consumption by grasshoppers: (1) cases obtained by asking a group of experts to solve representative hypothetical problems and (2) a numeric model of rangeland ecosystems. These knowledge sources are integrated through the technique of model-based adaptation, in which case-based reasoning is used to find an approximate solution, and the model is used to adapt this approximate solution into a more precise solution. CARMA has been used in Wyoming counties since 1996. The combination of a simple interface, flexible control strategy, and integration of multiple knowledge sources makes CARMA accessible to inexperienced users and capable of producing advice comparable to that produced by human experts.

Figure 1 illustrates grasshopper infestation densities in the western United States during 2000, a fairly typical year. In years of heavy infestation, grasshopper densities and economic losses might be much higher. For example, during the 1986 to 1987 outbreak, over 20 million acres of rangeland were treated for grasshoppers in the western United States at a cost of more than $75 million. In Wyoming, the estimated total annual loss to grasshoppers is roughly $19 million. The southeastern quadrant of the state is particularly prone to grasshopper infestations, with significant areas of high-grasshopper densities in 30 of the last 34 years.

CARMA is an advisory system for rangeland grasshopper infestations that demonstrates how AI technology can deliver expert advice to compensate for cutbacks in public services. CARMA uses two knowledge sources for the key task of predicting forage consumption by grasshoppers: (1) cases obtained by asking a group of experts to solve representative hypothetical problems and (2) a numeric model of rangeland ecosystems. These knowledge sources are integrated through the technique of model-based adaptation, in which case-based reasoning is used to find an approximate solution, and the model is used to adapt this approximate solution into a more precise solution. CARMA has been used in Wyoming counties since 1996. The combination of a simple interface, flexible control strategy, and integration of multiple knowledge sources makes CARMA accessible to inexperienced users and capable of producing advice comparable to that produced by human experts. Moreover, because CARMA embodies diverse forms of expertise, it has been used in ways that its developers did not anticipate, including pest management research, development of industry strategies, and in-state and federal pest-management policy decisions.