continual learning
In plain wordsThe ability of an AI to keep learning new things over time from a stream of experience, without forgetting what it already knew and without being retrained from scratch.
- 1
Machine learning
The study and practice of training a model on a non-stationary sequence of tasks or data distributions, updating it as data arrive while retaining earlier abilities and, ideally, using old knowledge to learn new things faster. The field is organized around the stability–plasticity dilemmastability–plasticity dilemma, n.: The basic tension any learning system faces: it must change enough to learn new things, yet stay stable enough not to lose what it already knows.: too little change and the learner cannot acquire new skills (loss of plasticityloss of plasticity, n.: The gradual loss of a neural network's ability to learn new things after long training on changing tasks, as parts of the network go inactive or become stuck.); too much and it suffers catastrophic forgettingcatastrophic forgetting, n.: When a neural network trained on something new gets suddenly and badly worse at what it could already do, because the new training rewrites the internal settings the old skills relied on.. Standard settings are task-, domain- and class-incremental learning. Method families are replay (storing or generating old examples), regularization (elastic weight consolidationelastic weight consolidation, n.: A way to help a neural network learn a new task without forgetting an old one, by making the settings that mattered most for the old task harder to change, as if tied down with springs., synaptic intelligence), and architectural approaches that add or isolate parameters (progressive networks, adaptersadapter, n.: A small trainable add-on slotted into a large frozen AI model to teach it a new task or style, like a plug adapter that lets one device work in many countries.). Evaluation tracks average accuracy, backward transfer (the effect of new learning on old tasks) and forward transfer.
Our continual learning setup adds a new product category each month and replays 5 percent of old data.
- 2
Language models and agentsemerging
The capacity, absent from most deployed large language modelslarge language model, n.: A computer program trained on enormous amounts of text to predict the words that come next. At vast scale, that simple skill lets it chat, answer questions, translate, summarize and write code. as of 2026, to update weights or durable memory from ongoing use — learning from a user's corrections, new documents or on-the-job experience — rather than relying on a fixed knowledge cutoffknowledge cutoff, n.: The date after which an AI model knows nothing from its training, because its training data stops there. Without a web search tool, it cannot know about later events. plus whatever fits in the context windowcontext window, n.: How much an AI model can read and keep in view at once: the instructions, the conversation so far and any attached documents. Text that does not fit has to be cut, summarized or left out.. Approaches under study include periodic fine-tuning, test-time trainingtest-time training, n.: Letting an AI model briefly update itself on each new problem it is given, learning from that problem's own clues before answering, instead of keeping its knowledge frozen after training., parametric memory modules and agent memoryagent memory, n.: The ways an AI assistant or agent keeps information — your preferences, past conversations, lessons from earlier tasks — and brings it back later, beyond what fits in a single conversation. stores; many researchers regard its absence as a key gap between current systems and human learners.
Origin
Mark Ring's 1994 dissertation was titled Continual Learning in Reinforcement Environments, and Sebastian Thrun and Tom Mitchell wrote on lifelong robot learning in 1995. The term revived with deep learning after the 2016–2017 work on elastic weight consolidation, and in 2025 continual learning became a prominent talking point about what large language models lack.
Usage note
Overlaps with online learningonline learning, n.: Learning from data that arrive one piece at a time, updating the model after each new example instead of training once on a fixed collection, as a spam filter does while emails stream in., which updates on a data stream but usually assumes a stationary target and does not measure forgetting. In product announcements 'continual learning' sometimes means only memory features that store notes, not weight updates.
Related
- catastrophic forgettingcatastrophic forgetting, n.: When a neural network trained on something new gets suddenly and badly worse at what it could already do, because the new training rewrites the internal settings the old skills relied on.
- stability–plasticity dilemmastability–plasticity dilemma, n.: The basic tension any learning system faces: it must change enough to learn new things, yet stay stable enough not to lose what it already knows.
- elastic weight consolidationelastic weight consolidation, n.: A way to help a neural network learn a new task without forgetting an old one, by making the settings that mattered most for the old task harder to change, as if tied down with springs.
- loss of plasticityloss of plasticity, n.: The gradual loss of a neural network's ability to learn new things after long training on changing tasks, as parts of the network go inactive or become stuck.
- online learningonline learning, n.: Learning from data that arrive one piece at a time, updating the model after each new example instead of training once on a fixed collection, as a spam filter does while emails stream in.
- test-time trainingtest-time training, n.: Letting an AI model briefly update itself on each new problem it is given, learning from that problem's own clues before answering, instead of keeping its knowledge frozen after training.
- agent memoryagent memory, n.: The ways an AI assistant or agent keeps information — your preferences, past conversations, lessons from earlier tasks — and brings it back later, beyond what fits in a single conversation.
- experience replayexperience replay, n.: Storing an AI agent's past experiences in a memory bank and repeatedly learning from random samples of them, rather than only from what just happened.
- online learningonline learning, n.: Learning from data that arrive one piece at a time, updating the model after each new example instead of training once on a fixed collection, as a spam filter does while emails stream in.
- transfer learningtransfer learning, n.: Reusing what a model learned on one task to get a head start on another — like a pianist picking up the organ faster than a beginner. It lets models learn new jobs from little data.
Sources
- Ring, M. B. (1994). Continual Learning in Reinforcement Environments. PhD dissertation, University of Texas at Austin.
- Kirkpatrick, J. et al. (2017). Overcoming catastrophic forgetting in neural networks. PNAS 114(13). arXiv:1612.00796. arxiv.org/abs/1612.00796
Mentioned in
agent memoryagent memory, n.: The ways an AI assistant or agent keeps information — your preferences, past conversations, lessons from earlier tasks — and brings it back later, beyond what fits in a single conversation., artificial general intelligenceartificial general intelligence, n.: AI that could learn and do almost any mental task a capable adult can, rather than excelling at one narrow job. Experts disagree on the exact bar and on when, or whether, it will arrive., catastrophic forgettingcatastrophic forgetting, n.: When a neural network trained on something new gets suddenly and badly worse at what it could already do, because the new training rewrites the internal settings the old skills relied on., continued pretrainingcontinued pretraining, n.: Taking an already-trained AI model and resuming its broad, read-everything training on new material — a specialty such as medicine or law, a new language, or fresher data — rather than starting from scratch., i.i.d.i.i.d., abbr.: Short for 'independent and identically distributed': every data point is produced the same way, and none influences any other — like repeated rolls of one fair die., learninglearning, n.: Getting better at something through experience. For a machine, it means changing its internal settings based on data so that it performs a task better than before., long-term memorylong-term memory, n.: Memory that lasts. For an AI agent, it means information saved outside the conversation — facts, past events, learned procedures — that can be looked up and used again days or months later., online learningonline learning, n.: Learning from data that arrive one piece at a time, updating the model after each new example instead of training once on a fixed collection, as a spam filter does while emails stream in., elastic weight consolidationelastic weight consolidation, n.: A way to help a neural network learn a new task without forgetting an old one, by making the settings that mattered most for the old task harder to change, as if tied down with springs., hypernetworkhypernetwork, n.: A neural network whose job is to produce the settings, or weights, of another neural network, so that one model can generate many specialized models on demand., loss of plasticityloss of plasticity, n.: The gradual loss of a neural network's ability to learn new things after long training on changing tasks, as parts of the network go inactive or become stuck., nested learningnested learning, n.: A Google research idea that treats an AI model as layers of learning processes updating at different speeds, with the aim of letting models keep learning without forgetting old skills., stability–plasticity dilemmastability–plasticity dilemma, n.: The basic tension any learning system faces: it must change enough to learn new things, yet stay stable enough not to lose what it already knows., test-time trainingtest-time training, n.: Letting an AI model briefly update itself on each new problem it is given, learning from that problem's own clues before answering, instead of keeping its knowledge frozen after training., Tom MitchellTom Mitchell, prop. n.: An American computer scientist who wrote a classic machine-learning textbook with the field's most quoted definition of learning, and founded the first university department devoted to machine learning.