CNR Author
operator:
and / or
Typology
operator:
and / or
Language
operator:
and / or
Date
operator:
and / or
Rights
operator:
and / or
2025
Conference article
Open Access
The Future of continual learning in the era of foundation models: three key directions
Bell J., Quarantiello L., Coleman E. N., Li L., Li M., Madeddu M., Piccoli E., Lomonaco V.
Continual learning--the ability to acquire, retain, and refine knowledge over time--has always been fundamental to intelligence, both human and artificial. Historically, different AI paradigms have acknowledged this need, albeit with varying priorities: early expert and production systems focused on incremental knowledge consolidation, while reinforcement learning emphasised dynamic adaptation. With the rise of deep learning, deep continual learning has primarily focused on learning robust and reusable representations over time to solve sequences of increasingly complex tasks. However, the emergence of Large Language Models (LLMs) and foundation models has raised the question: Do we still need continual learning when centralised, monolithic models can tackle diverse tasks with access to internet-scale knowledge? We argue that continual learning remains essential for three key reasons: (i) continual pre-training is still necessary to ensure foundation models remain up to date, mitigating knowledge staleness and distribution shifts while integrating new information; (ii) continual fine-tuning enables models to specialise and personalise, adapting to domain-specific tasks, user preferences, and real-world constraints without full retraining, avoiding the need for computationally expensive long context-windows; (iii) continual compositionality offers a scalable and modular approach to intelligence, enabling the orchestration of foundation models and agents to be dynamically composed, recombined, and adapted. While continual pre-training and fine-tuning are explored as niche research directions, we argue it is continual compositionality that will mark the rebirth of continual learning. The future of AI will not be defined by a single static model but by an ecosystem of continually evolving and interacting models, making continual learning more relevant than ever.Source: CEUR WORKSHOP PROCEEDINGS, vol. 4074, pp. 525-548. University of Pisa/Scuola Normale Superiore, Pisa, Italy, 9-102025
Bell J., Quarantiello L., Coleman E. N., Li L., Li M., Madeddu M., Piccoli E., Lomonaco V.
Continual learning--the ability to acquire, retain, and refine knowledge over time--has always been fundamental to intelligence, both human and artificial. Historically, different AI paradigms have acknowledged this need, albeit with varying priorities: early expert and production systems focused on incremental knowledge consolidation, while reinforcement learning emphasised dynamic adaptation. With the rise of deep learning, deep continual learning has primarily focused on learning robust and reusable representations over time to solve sequences of increasingly complex tasks. However, the emergence of Large Language Models (LLMs) and foundation models has raised the question: Do we still need continual learning when centralised, monolithic models can tackle diverse tasks with access to internet-scale knowledge? We argue that continual learning remains essential for three key reasons: (i) continual pre-training is still necessary to ensure foundation models remain up to date, mitigating knowledge staleness and distribution shifts while integrating new information; (ii) continual fine-tuning enables models to specialise and personalise, adapting to domain-specific tasks, user preferences, and real-world constraints without full retraining, avoiding the need for computationally expensive long context-windows; (iii) continual compositionality offers a scalable and modular approach to intelligence, enabling the orchestration of foundation models and agents to be dynamically composed, recombined, and adapted. While continual pre-training and fine-tuning are explored as niche research directions, we argue it is continual compositionality that will mark the rebirth of continual learning. The future of AI will not be defined by a single static model but by an ecosystem of continually evolving and interacting models, making continual learning more relevant than ever.Source: CEUR WORKSHOP PROCEEDINGS, vol. 4074, pp. 525-548. University of Pisa/Scuola Normale Superiore, Pisa, Italy, 9-102025
See at:
ceur-ws.org 
| CNR IRIS | CNR IRIS 
2025
Conference article
Open Access
A compositional paradigm for foundation models: towards smarter robotic agents
Quarantiello L., Piccoli E., Bell J., Li M., Carfì G., Coleman E. N., Gramaglia G., Li L., Madeddu M., Testa I., Lomonaco V.
The birth of Foundation Models brought unprecedented results in a wide range of tasks, from language to vision, to robotic control. These models are able to process huge quantities of data, and can extract and develop rich representations, which can be employed across different domains and modalities. However, they still have issues in adapting to dynamic, real-world scenarios without retraining the entire model from scratch. In this work, we propose the application of Continual Learning and Compositionality principles to foster the development of more flexible, efficient and smart AI solutions.DOI: 10.5281/zenodo.17629900
DOI: 10.5281/zenodo.17629899
DOI: 10.48550/arxiv.2510.18608
Metrics:
Quarantiello L., Piccoli E., Bell J., Li M., Carfì G., Coleman E. N., Gramaglia G., Li L., Madeddu M., Testa I., Lomonaco V.
The birth of Foundation Models brought unprecedented results in a wide range of tasks, from language to vision, to robotic control. These models are able to process huge quantities of data, and can extract and develop rich representations, which can be employed across different domains and modalities. However, they still have issues in adapting to dynamic, real-world scenarios without retraining the entire model from scratch. In this work, we propose the application of Continual Learning and Compositionality principles to foster the development of more flexible, efficient and smart AI solutions.DOI: 10.5281/zenodo.17629900
DOI: 10.5281/zenodo.17629899
DOI: 10.48550/arxiv.2510.18608
Metrics:
See at:
arXiv.org e-Print Archive | CNR IRIS | proceedings.i-rim.it | doi.org | ZENODO | ZENODO | CNR IRIS
2025
Conference article
Open Access
Adaptive AI-based decentralized resource management in the cloud-edge continuum
Li L., Bell J., Coppola M., Lomonaco V.
In the Cloud-Edge Continuum, dynamic infrastructure change and variable workloads complicate efficient resource management. Centralized methods can struggle to adapt, whilst purely decentralized policies lack global oversight. This paper proposes a hybrid framework using Graph Neural Network (GNN) embeddings and collaborative multi-agent reinforcement learning (MARL). Local agents handle neighbourhood-level decisions, and a global orchestrator coordinates system-wide. This work contributes to decentralized application placement strategies with centralized oversight, GNN integration and collaborative MARL for efficient, adaptive and scalable resource management.DOI: 10.1109/pdp66500.2025.00053
Project(s): CHARITY
Metrics:
Li L., Bell J., Coppola M., Lomonaco V.
In the Cloud-Edge Continuum, dynamic infrastructure change and variable workloads complicate efficient resource management. Centralized methods can struggle to adapt, whilst purely decentralized policies lack global oversight. This paper proposes a hybrid framework using Graph Neural Network (GNN) embeddings and collaborative multi-agent reinforcement learning (MARL). Local agents handle neighbourhood-level decisions, and a global orchestrator coordinates system-wide. This work contributes to decentralized application placement strategies with centralized oversight, GNN integration and collaborative MARL for efficient, adaptive and scalable resource management.DOI: 10.1109/pdp66500.2025.00053
Project(s): CHARITY
Metrics:
See at:
CNR IRIS | ieeexplore.ieee.org | CNR IRIS | CNR IRIS
2024
Conference article
Open Access
Continual policy distillation of reinforcement learning-based controllers for soft robotic in-hand manipulation
Li L., Donato E., Lomonaco V., Falotico E.
Dexterous manipulation, often facilitated by multi-fingered robotic hands, holds solid impact for real-world ap-plications. Soft robotic hands, due to their compliant nature, offer flexibility and adaptability during object grasping and manipulation. Yet, benefits come with challenges, particularly in the control development for finger coordination. Reinforce-ment Learning (RL) can be employed to train object-specific in-hand manipulation policies, but limiting adaptability and generalizability. We introduce a Continual Policy Distillation (CPD) framework to acquire a versatile controller for in-hand manipulation, to rotate different objects in shape and size within a four-fingered soft gripper. The framework leverages Policy Distillation (PD) to transfer knowledge from expert policies to a continually evolving student policy network. Exemplar-based rehearsal methods are then integrated to mitigate catastrophic forgetting and enhance generalization. The performance of the CPD framework over various replay strategies demonstrates its effectiveness in consolidating knowledge from multiple experts and achieving versatile and adaptive behaviours for in-hand manipulation tasks.DOI: 10.1109/robosoft60065.2024.10522027
DOI: 10.48550/arxiv.2404.04219
Project(s): PROBOSCIS
Metrics:
Li L., Donato E., Lomonaco V., Falotico E.
Dexterous manipulation, often facilitated by multi-fingered robotic hands, holds solid impact for real-world ap-plications. Soft robotic hands, due to their compliant nature, offer flexibility and adaptability during object grasping and manipulation. Yet, benefits come with challenges, particularly in the control development for finger coordination. Reinforce-ment Learning (RL) can be employed to train object-specific in-hand manipulation policies, but limiting adaptability and generalizability. We introduce a Continual Policy Distillation (CPD) framework to acquire a versatile controller for in-hand manipulation, to rotate different objects in shape and size within a four-fingered soft gripper. The framework leverages Policy Distillation (PD) to transfer knowledge from expert policies to a continually evolving student policy network. Exemplar-based rehearsal methods are then integrated to mitigate catastrophic forgetting and enhance generalization. The performance of the CPD framework over various replay strategies demonstrates its effectiveness in consolidating knowledge from multiple experts and achieving versatile and adaptive behaviours for in-hand manipulation tasks.DOI: 10.1109/robosoft60065.2024.10522027
DOI: 10.48550/arxiv.2404.04219
Project(s): PROBOSCIS
Metrics:
See at:
arXiv.org e-Print Archive | CNR IRIS | ieeexplore.ieee.org | doi.org | doi.org | Archivio della ricerca della Scuola Superiore Sant'Anna | Archivio della Ricerca - Università di Pisa | CNR IRIS | CNR IRIS
2024
Conference article
Open Access
Calibration of continual learning models
Li L., Piccoli E., Cossu A., Bacciu D., Lomonaco V.
Continual Learning (CL) focuses on maximizing the predictive performance of a model across a non-stationary stream of data. Unfortunately, CL models tend to forget previous knowledge, thus often underperforming when compared with an offline model trained jointly on the entire data stream. Given that any CL model will eventually make mistakes, it is of crucial importance to build calibrated CL models: models that can reliably tell their confidence when making a prediction. Model calibration is an active research topic in machine learning, yet to be properly investigated in CL. We provide the first empirical study of the behavior of calibration approaches in CL, showing that CL strategies do not inherently learn calibrated models. To mitigate this issue, we design a continual calibration approach that improves the performance of post-processing calibration methods over a wide range of different benchmarks and CL strategies. CL does not necessarily need perfect predictive models, but rather it can benefit from reliable predictive models. We believe our study on continual calibration represents a first step towards this direction.Source: IEEE COMPUTER SOCIETY CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION WORKSHOPS, vol. 24, pp. 4160-4169. Seattle, USA, 2024
DOI: 10.1109/cvprw63382.2024.00419
Project(s): EMERGE, Future Artificial Intelligence Research
Metrics:
Li L., Piccoli E., Cossu A., Bacciu D., Lomonaco V.
Continual Learning (CL) focuses on maximizing the predictive performance of a model across a non-stationary stream of data. Unfortunately, CL models tend to forget previous knowledge, thus often underperforming when compared with an offline model trained jointly on the entire data stream. Given that any CL model will eventually make mistakes, it is of crucial importance to build calibrated CL models: models that can reliably tell their confidence when making a prediction. Model calibration is an active research topic in machine learning, yet to be properly investigated in CL. We provide the first empirical study of the behavior of calibration approaches in CL, showing that CL strategies do not inherently learn calibrated models. To mitigate this issue, we design a continual calibration approach that improves the performance of post-processing calibration methods over a wide range of different benchmarks and CL strategies. CL does not necessarily need perfect predictive models, but rather it can benefit from reliable predictive models. We believe our study on continual calibration represents a first step towards this direction.Source: IEEE COMPUTER SOCIETY CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION WORKSHOPS, vol. 24, pp. 4160-4169. Seattle, USA, 2024
DOI: 10.1109/cvprw63382.2024.00419
Project(s): EMERGE
, Future Artificial Intelligence ResearchMetrics:
See at:
CNR IRIS | ieeexplore.ieee.org | CNR IRIS | CNR IRIS | CNR IRIS