The paper introduces Entropic Control for Homeostatic Optimization (ECHO), a closed‑loop framework that mitigates catastrophic forgetting in supervised fine‑tuning of large language models. By decomposing KL divergence, the authors identify output entropy as a key observable, allowing real‑time monitoring of forgetting without costly reference‑model inference. ECHO uses an elastic controller to adjust replay‑data gradient weights based on entropy deviations, and experiments on 0.6B–9B parameter models show it improves the plasticity‑stability trade‑off and remains robust across different data mixing ratios.
arXiv:2608. 11690v1 Announce Type: new Abstract: Continual learning must absorb new tasks without erasing old ones, and replay---mixing a small buffer of past examples into current training---is among the most effective remedies for catastrophic forgetting.
arXiv:2606. 08452v1 Announce Type: new Abstract: In many real-world settings, data streams are nonstationary and arrive sequentially, requiring learning systems to adapt continuously without retraining from scratch.
The paper introduces Energy-Based Generative Replay (EB-Gen) for Equilibrium Propagation (EqProp) trained Hopfield Networks, enabling continual learning without external data buffers or architectural overhead. EB-Gen leverages the network’s intrinsic generative dynamics to produce pseudo‑samples from a frozen copy of the model, allowing internal rehearsal. Experiments on split MNIST and split CIFAR‑10 show that EB-Gen improves stability and plasticity, achieving 75 % and 64 % average accuracy respectively, outperforming several baseline methods.
arXiv:2603. 11395v3 Announce Type: replace-cross Abstract: Continual reinforcement learning challenges agents to acquire new skills while retaining previously learned ones with the goal of improving performance in both past and future tasks.
The paper introduces SVC, a parameter‑efficient continual‑learning method for large language models that uses singular‑vector channels as the unit for balancing plasticity and stability. By estimating each channel’s adaptation benefit and forgetting cost from domain‑specific and general‑domain data, SVC selectively updates only the most beneficial channels, employing knee‑based cost screening, Pareto‑front filtering, and Otsu thresholding. Experiments on four LLM families and eight downstream tasks show that SVC preserves pretrained capabilities while achieving strong performance compared to existing PEFT baselines.
We introduce CMP (Cognitive Memory Primitive), a continual-learning architecture that repre? sents inputs as sparse relational codes, stores them in a two-tier competitive memory, and learns through local updates without end-to-end backpropagation through its feature-generating system.
arXiv:2609.38768v1 Announce Type: cross Abstract: Prior work has shown that neural networks exhibit implicit biases toward low-complexity structure (e.g., spectral bias), memorization dynamics, and c...
arXiv:2608. 03660v1 Announce Type: new Abstract: Reinforcement fine-tuning (RFT) is widely believed to inherently resist catastrophic forgetting in continual post-training of multimodal large language models.
arXiv:2607. 26523v1 Announce Type: new Abstract: We introduce CMP (Cognitive Memory Primitive), a continual-learning architecture that repre?
The paper introduces a Fisher-guided submodular data selection method for continual pre‑training of large language models, addressing the forgetting problem that arises when a target‑domain corpus overwrites pretrained knowledge. By decomposing candidate gradients into anchor and frontier components and optimizing a log‑determinant submodular objective, the method selects data that both improves target‑domain performance and limits forgetting. Experiments on TinyLlama‑1.1B and Llama‑3.1‑8B show that the selector achieves superior adaptation and forgetting control while using ten times fewer tokens than traditional replay strategies.
The paper investigates catastrophic forgetting in the scientific foundation model Walrus-1.3B during a two‑stage sim‑to‑real adaptation. Finetuning first on simulated flow around a cylinder quickly degrades performance on the original pretraining data, with one‑step errors rising up to 35× after 20 epochs, and further adaptation to experimental data worsens this degradation. The authors show that replaying as little as 1% of the simulated data during real‑data finetuning largely restores simulation performance while still adapting effectively to real dynamics.
Prior work has shown that neural networks exhibit implicit biases toward low-complexity structure (e.g., spectral bias), memorization dynamics, and compression-like effects during training, but a unif...