Continual Learning

TL;DR for operators Fleets do not fail only because they forget. They also fail because they remember the wrong thing at the wrong time. That is the practical point of COMAD, a framework for continual offline multi-agent reinforcement learning proposed in Offline Multi-agent Continual Cooperation via Skill Partition and Reuse.1 The paper studies agents that must learn from a stream of offline datasets: first one cooperative task, then another, then another, without interactive trial-and-error and without assuming the required coordination skills stay fixed. That setting is awkward, which is why it is useful. Real deployed systems rarely receive the courtesy of a clean, stationary benchmark and a polite email before the operating conditions change. ...

Updating a model sounds harmless until the old parts of the system start reading the new representations incorrectly. That is the less theatrical version of catastrophic forgetting. Not the dramatic story where a neural network “forgets everything” like a distracted intern. The more useful story is quieter: a deployed AI system adapts its backbone to new data, the feature space shifts, and classifiers trained for earlier tasks are left calibrated to yesterday’s geometry. ...

Forgetting That Never Happened: The Shallow Alignment Trap Forgetfulness is an expensive diagnosis. When an internal AI system performs well on last month’s support taxonomy, then underperforms after being fine-tuned on this month’s compliance cases, the obvious story is simple: the model forgot. That story usually triggers an equally obvious response: replay old data, retrain more broadly, freeze more parameters, or panic politely in a meeting while calling it “model lifecycle management.” ...

Avatars are easy to make expressive once. That is the boring version of the problem. Give a motion model enough examples of sad walking, angry gesturing, or excited dancing, and it can learn the broad association between text and motion. The harder problem starts later, after the product has already shipped. A game studio adds a new combat animation pack. A VR training company expands from office scenarios to emergency response. A digital-human platform moves from daily-life gestures into sports, performance, musical instruments, and acrobatics. Suddenly “sad” is no longer just a lowered head during walking. It must become a lowered head while jogging, a constrained body during performance, or a professional movement pattern inside a sport. ...

Training an AI model is often sold like a kitchen renovation: add more data, add reinforcement learning, install the shiny reasoning countertop, and suddenly the whole thing looks expensive enough to be intelligent. This paper is useful because it ruins that brochure. The authors of Atomic Skills are the Prerequisite: When Reinforcement Learning Synthesizes Compositional Reasoning, and When It Only Amplifies ask a deceptively simple question: does reinforcement learning create new reasoning ability, or does it only increase the probability of behaviors the model could already produce?1 Their answer is not the clean slogan either camp wants. RL can synthesize new compositional reasoning, but only when the model has already learned the right underlying atomic skills. Without that foundation, RL mostly polishes whatever behavior already exists. Sometimes that is reasoning. Sometimes it is just a better-trained shortcut wearing a lab coat. ...

Image generators fail in very ordinary ways. A prompt asks for a green banana and a blue vase. The model gives you something banana-adjacent, vase-adjacent, and chromatically negotiable. A designer asks for a bowl containing a pizza. The model places the pizza beside the bowl, halfway inside the bowl, or in a bowl-like universe where geometry has apparently resigned. A product team then does the usual dance: collect bad outputs, ask users what they preferred, curate examples, fine-tune later, and call the whole thing “continuous improvement” because the spreadsheet had a date column. ...

Catalogs change. Defect categories change. Fraud patterns change. Document types change. The model, unfortunately, often reacts like an employee who learns the new product line and immediately forgets where the old shelves are. That is the everyday problem behind Class-Incremental Learning (CIL): a model must learn new classes over time while still recognizing old ones. The difficult part is not merely adding output labels. It is keeping the feature extractor from being rewritten by the latest task until yesterday’s knowledge becomes decorative archaeology. ...

TL;DR for operators Model composition usually sounds harmless until someone asks the obvious production question: “Can we remove that client-specific update without retraining the whole thing?” At that point, many elegant AI stacks quietly become sedimentary rock. The MDM-OC paper proposes a cleaner model lifecycle: keep a shared base model, express every fine-tuned specialist as a task delta, orthogonalize those deltas so they interfere less, merge them with tuned coefficients, and subtract a selected delta later when a capability, customer, or data source needs to be removed.1 The important claim is not “we found another averaging recipe.” The claim is that model updates can be treated as separable components in parameter space. ...