We introduce CMP (Cognitive Memory Primitive), an architecture that represents inputs as sparse relational codes, stores them in a two-tier competitive memory, and learns entirely through local, gradient-free updates, with no backpropagation anywhere in the network. We use this architecture to test a specific hypothesis: that catastrophic forgetting, usually treated as a training-time defect to be patched with replay or regularization, is instead a structural consequence of how backpropagation assigns credit and that a learning rule that is local and sparse by construction should resist it without a patch. On a controlled domain-incremental protocol across 15 text domains, three-seed replicated, CMP's backward transfer is 15-19x better than a matched-size Transformer trained with online EWC, and the result survives a domain-order control (reported as a range, +0.24 to +0.44, rather than a single figure). We report this alongside a real, substantial accuracy gap versus the Transformer baseline, a null result on a recognized vision benchmark, and a diagnosed, unresolved failure attempting to combine this architecture with a separate mechanism that improves raw accuracy, disclosed because an honest negative result is more useful than an omitted one. The central claim is narrow and falsifiable: local, sparse, non-backpropagation learning measurably resists catastrophic forgetting better than backpropagation with its standard fix, under conditions we state precisely.
The results suggest that the combination of sparse representations, local learning, and persistent memory is a promising direction for continual learning, while motivating further investigation into the respective roles of learning rules, representations, and architectural design in mitigating catastrophic forgetting.
Current mainstream deep learning techniques exhibit an over-reliance on extensive training data and a lack of adaptability to the dynamic world, marking a considerable disparity from human intelligence. To bridge this gap, Few-Shot Class-Incremental Learning(FSCIL) has emerged, focusing on continuous learning of new categories with limited samples without forgetting old knowledge. Existing FSCIL studies typically use a single model to learn knowledge across all sessions, inevitably leading to the stability–plasticity dilemma. Unlike machines that usually consolidate all categories into a single parameter space, cortical memory organization suggests that different types of knowledge can be distributed and organized across specialized cortical regions. Inspired by this organization principle, our paper aims to develop a method that learns independent models for each session. It can inherently prevent catastrophic forgetting. During the testing stage, our method integrates Uncertainty Quantification (UQ) for model deployment. Our method provides a fresh viewpoint for FSCIL and demonstrates the state-of-the-art performance on CIFAR-100 and mini-ImageNet datasets.
Renye Zhang, Yi Yin, Jinghua Zhang· Entropy· 0 citations
AI Engram (Kwon et al., 2026) formalizes the four engram criteria of neuroscience as a constrained inverse problem in weight space and solves it closed-form: concept-specific memory traces become linear objects that can be extracted once and combined arithmetically. Appendix F states the Compositional Memory States Hypothesis: edited models live on"a commutative manifold where the integration of A and B reaches a consistent equilibrium regardless of the learning sequence."The evidence base is single and paired edits -- in materials terms, single-cycle tests, in which fatigue accumulation is structurally invisible. Whether the hypothesis holds under sequential load is exactly the"temporal dynamics"question the paper defers to future work. We run that test on the authors'own reference implementation, at their reported best edit strength (TOFU alpha=0.6, a choice favoring the linearity hypothesis), with pre-registered predictions, across three model charges (two vendors, two architecture families). Four findings replicate across all three: (1) zero-shot composition and sequential re-calibrated editing diverge by 61-71% of the edit magnitude; (2) cut order is not interchangeable, and the effect scales with concept overlap -- in one charge the order of cutting two Paris landmarks decides whether an uninvolved third concept survives; (3) the survivors'layer-input covariances -- the method's own sufficient statistics, read as strain gauges -- drift monotonically with every further cut, in every surviving concept, in every charge; (4) erased knowledge partially returns under subsequent unrelated cuts. Appendix F's commutative-manifold hypothesis is thereby falsified for sequential editing; the single-edit results of the original paper are untouched. For unlearning-as-compliance: erasure certified today does not certify the artifact after its next edit.
This paper demonstrates that standard activations -- whether piecewise-linear (ReLU, PReLU, Hardtanh) or smooth (SiLU, Sigmoid, Tanh, GELU) -- are in fact instances of a single Threshold Gating (TG) primitive, and proposes a'Minimal Branch Theorem' which relates the minimum number of required branches in the authors' primitive to the trainability of general deep neural networks.
Muhammad Sabih, Frank Hannig, Jürgen Teich· 0 citations
Humans retain memories of individual experiences for a lifetime, an ability attributed to a complementary learning system in which a fast process encodes episodes and a slow process integrates them into semantic knowledge. In classical Hebbian models such as Hopfield networks, memory traces are superposed in shared weights. This makes learning naturally continual but causes strong interference among correlated memories, a failure that reappears as catastrophic forgetting in deep networks. Here we use a large language model as a model system for continual episodic memory, with its pretrained weights supplying the semantic context in which new episodes are embedded. Fast learning is implemented by a hippocampus-like module that assigns each episode to a dedicated, extremely sparse low-rank adapter; competitive gating then selects among these separated traces during recall. Across streams of up to 1,000 factual and autobiographical episodes, each adapter requires only 2–3 parameters per token while preserving excellent recall. An internal retrieval-augmented generation mechanism reconstructs the selected episode in context and supports high-accuracy question answering over stored memories. Finally, slow cortical consolidation is modeled by fine-tuning the base weights through batch replay, enabling reconstruction and direct question answering without episodic adapters. Together, fast storage and slow consolidation implement both components of a complementary learning system within a single language model, yielding a neural-network model that stores, recalls, and consolidates naturalistic episodic memories, thereby capturing key functional features of human memory.
Xu Pan, Ely Hahami, R. Siegelmann et al.· bioRxiv· 0 citations
MemDefrag, a training-free and model-agnostic framework that uses a middle-layer tracing signal to conduct memory defragmentation (rank, reorder, and filter memories), and applies an informativeness-guided proportional forgetting mechanism once capacity is exceeded, is proposed.