Catastrophic forgetting
Catastrophic forgetting is the substantial loss of previously learned information or capabilities when a model learns new data or tasks.
Building Specialized AI Systems
Catastrophic forgetting is the abrupt loss of previously learned capabilities when a model is adapted to new data or tasks.
- How it is used
- Frames the stability risk when choosing how aggressively to adapt pretrained models for specialized tasks.
- How deep it goes
- Explains forgetting through all-layer updates, frozen-layer mitigation, and domain-data mixing without detailing internal mechanisms.
- Examples
- Uses legal-English specialization and a clinical masked-language-model corpus for domain adaptation.
- What you can do after
- Evaluate fine-tuning choices by balancing specialization, compute cost, and preservation of pretrained capabilities.
Read how Building Specialized AI Systems explains catastrophic forgetting →
Reinforcement Learning from Human Feedback
Catastrophic forgetting is the rapid loss of previously learned capabilities or information when a model is trained on new objectives or data.
- How it is used
- Uses forgetting to compare prior-task retention during sequential RL and supervised fine-tuning.
- How deep it goes
- Derives forward- and reverse-KL views of the objectives and examines sampling and gradient effects on retention.
- Examples
- Illustrates old and new policy modes and compares forgetting across RL and SFT variants.
- What you can do after
- Diagnoses potential forgetting by measuring policy KL drift on new-task inputs, while treating it as an empirical estimate.
Read how Reinforcement Learning from Human Feedback explains catastrophic forgetting →
The Art of AI Product Development
Catastrophic forgetting is the loss of previously learned knowledge when a model is trained on new tasks or data.
- How it is used
- Uses evaluation on older tasks to decide whether fine-tuning has damaged retained knowledge.
- How deep it goes
- Defines the failure, identifies overlapping representations as a mechanism, and outlines evaluation and prevention decisions.
- Examples
- Uses acronym resolution, industry-term definition, benchmark regression, and medical, legal, and compliance applications.
- What you can do after
- Diagnose forgetting by comparing performance on previously learned tasks before and during new-task training.
Read how The Art of AI Product Development explains catastrophic forgetting →
LLM Customization and Fine-Tuning
The loss of previously learned capabilities when a model is fine-tuned on narrower data, as updates overwrite parameters supporting older skills.
- How it is used
- Guides risk management for full supervised fine-tuning and incremental retraining.
- How deep it goes
- Explains the overwrite mechanism and gives concrete prevention, evaluation, and deployment steps.
- Examples
- Uses general-purpose golden sets and reported self-distillation benchmark comparisons to illustrate evaluation and alternatives.
- What you can do after
- Diagnose capability loss and safety regression by comparing general-purpose and safety tests before and after fine-tuning.
Read how LLM Customization and Fine-Tuning explains catastrophic forgetting →
Generative AI in Action
Catastrophic forgetting is the loss of previously learned knowledge or capabilities when a model is trained on new data or tasks.
- How it is used
- Frames model adaptation as a balance between task specialization and retaining pretrained knowledge and reasoning.
- How deep it goes
- Explains parameter overwriting and PEFT mitigation, while deferring detailed adaptation-technique trade-offs.
- Examples
- Uses sentiment analysis, spam detection, question answering, summarization, and translation as adaptation examples.
- What you can do after
- Apply it via PEFT by updating a small parameter subset while keeping most pretrained parameters fixed.
Read how Generative AI in Action explains catastrophic forgetting →
Also covered in
- How Large Language Models Work
- Rearchitecting LLMsIn early access (MEAP)
- Machines that ThinkIn early access (MEAP)
- Building Reliable AI Systems
What to know first
- fine-tuningFine-tuning is the continued training of a pretrained model on targeted data or objectives to adapt its parameters or behavior to a task, domain, or use.
- language modelA language model is a computational model that represents patterns in linguistic sequences to assign probabilities, predict, generate, or evaluate language.
- foundation modelA foundation model is a broadly trained machine-learning model whose learned capabilities can be reused or adapted for a range of downstream tasks.
- LlamaLlama is a family of large language models developed by Meta.
Related concepts
- parameter-efficient fine-tuningaddressed by · Parameter-efficient fine-tuning adapts a pretrained model by training a limited set of existing or added parameters while leaving most of the model unchanged.
- Kullback–Leibler divergencemeasured by · Kullback–Leibler divergence is a nonnegative, generally asymmetric measure of the difference between probability distributions, defined as the expected log probability ratio under one distribution.
- low-rank adaptationaddressed by · Low-rank adaptation (LoRA) is a parameter-efficient fine-tuning technique that adapts a model by learning low-rank updates to its weights, typically while keeping most original weights fixed.
- supervised fine-tuningproduced by · Supervised fine-tuning is further training a pretrained model on labeled input-output examples to adapt its behavior to desired outputs.
- forgettingkind of · Forgetting is the reduction or removal of retained information or older observations from memory or their influence on subsequent estimates.
- layer-wise fine-tuningaddressed by · Layer-wise fine-tuning is a model-adaptation method that sequentially trains one neural-network layer or layer group at a time while keeping the other layers frozen.
Questions about catastrophic forgetting
What is catastrophic forgetting?
Catastrophic forgetting is the substantial loss of previously learned information or capabilities when a model learns new data or tasks.
Which Manning books explain catastrophic forgetting?
Building Specialized AI Systems, Reinforcement Learning from Human Feedback, The Art of AI Product Development, LLM Customization and Fine-Tuning and Generative AI in Action.
What should I know before learning catastrophic forgetting?
Manning books that teach catastrophic forgetting usually introduce fine-tuning, language model, foundation model and Llama first.
What is catastrophic forgetting a kind of?
Catastrophic forgetting is a kind of forgetting. Forgetting is the reduction or removal of retained information or older observations from memory or their influence on subsequent estimates.
These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.




