aiqbench analysis
This arXiv paper proposes a Complementary Learning Systems-inspired architecture in which a short-term memory offloads generalization to a slow, long-term memory, aiming to resolve the tension between few-shot generalization and continual adaptation. Results on image classification are modest: better resistance to forgetting and stronger few-shot generalization than baselines, but slightly lower overall continual-learning accuracy than existing methods, with the authors pointing to the long-term memory's representations as the bottleneck. For ML readers, it is a early-stage research contribution to continual learning rather than a practical advance, and its main value is the design idea and the open problem of replaying short-term memories to long-term memory.
Episodic looking - using reinforcement learning to bias perception enhances few-shot and continual learning with protection from forgetting
arXiv:2610.38577v1 Announce Type: cross Abstract: There is a fundamental contradiction between the desire for models that can generalize, and models that can continuously adapt to recent input with changing statistics. These conditions, captured in few-shot and continual learning settings, can bias or disrupt learned representations. This harms generalization to a broader context, leaving the model unable to overcome input variance. This paper proposes a solution based on the Complementary Learning System architecture (CLS), in which a Short-Term Memory (STM) delegates generalization to a slow, statistical Long-Term Memory (LTM). Delegation is achieved by working in a space of robust, generalized perceptions produced by the LTM. With this separation, the STM can focus on memorizing recent input and avoiding interference between memories. We show that on an image classification task, the resulting CLS model is more resistant to forgetting in a domain-incremental continual learning setting than a baseline ResNet LTM, and generalizes more effectively in a few-shot setting than other continual learners. In the continual setting, overall accuracy is slightly worse than existing continual learning models, but with a different mix of performance characteristics. Results analysis suggests that the model is limited by the representations and representational plasticity of the LTM. In future work, we will seek to address this by replaying STM memories to the LTM.
Abstract:There is a fundamental contradiction between the desire for models that can generalize, and models that can continuously adapt to recent input with changing statistics. These conditions, captured in few-shot and continual learning settings, can bias or disrupt learned representations. This harms generalization to a broader context, leaving the model unable to overcome input variance. This paper proposes a solution based on the Complementary Learning System architecture (CLS), in which a Short-Term Memory (STM) delegates generalization to a slow, statistical Long-Term Memory (LTM). Delegation is achieved by working in a space of robust, generalized perceptions produced by the LTM. With this separation, the STM can focus on memorizing recent input and avoiding interference between memories. We show that on an image classification task, the resulting CLS model is more resistant to forgetting in a domain-incremental continual learning setting than a baseline ResNet LTM, and generalizes more effectively in a few-shot setting than other continual learners. In the continual setting, overall accuracy is slightly worse than existing continual learning models, but with a different mix of performance characteristics. Results analysis suggests that the model is limited by the representations and representational plasticity of the LTM. In future work, we will seek to address this by replaying STM memories to the LTM.
Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
I.2.6; I.2.6; I.5.1
arXiv:2610.38577 [cs.LG]
(or arXiv:2610.38577v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2610.38577 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 310 words
Related coverage
Omni-Perception Policy Optimization for Multimodal Emotion Reasoning
arXiv · Jul 21, 2026
When Scaling Fails: Mitigating Audio Perception Decay of LALMs via Multi-Step Perception-Aware Reasoning
arXiv · Sep 22, 2026
ACTIVE-o3: Empowering MLLMs with Active Perception via Pure Reinforcement Learning
arXiv · Jun 9, 2026
PGTT: Phase-Guided Terrain Traversal for Perceptive Legged Locomotion
arXiv · Jul 23, 2026