Sparsity, Superposition, and Forgetting: A Mechanistic Study of Representation Retention in Continual Learning

• URL: http://arxiv.org/abs/2606.20431v1
• Date: Thu, 18 Jun 2026 16:10:40 GMT
• Title: Sparsity, Superposition, and Forgetting: A Mechanistic Study of Representation Retention in Continual Learning
• Abstract summary: Continual learning systems often forget previously acquired knowledge.<n>We present a controlled, toy-world framework that makes forgetting mechanisms observable and testable.
• Score: 6.113106953880908
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Continual learning (CL) systems often forget previously acquired knowledge, yet the mechanisms driving forgetting remain hard to isolate in practice because real datasets entangle many factors. We present a controlled, toy-world framework that makes these mechanisms observable and testable. Using a synthetic generator-separator pipeline, we define ground-truth latent features, build tasks with tunable sparsity and overlap, and introduce measurable quantities for representation strength and superposition (directional overlap among features). We then study retention dynamics-the temporal change of representation strength by fitting sparse dynamical relations (via SINDy) between retention, superposition, and exposure history. A complementary task-level analysis based on effective rank characterizes how representational capacity is allocated across tasks. Our controlled experiments yield three takeaways. (1) Superposition tends to increase over time with transient dips at task boundaries, suggesting boundary-specific interference rather than steady drift. (2) Higher feature sparsity induces more superposition yet does not inevitably cause forgetting; when representations remain strong, forgetting can be reduced despite overlap. (3) Task-level effective rank grows with sparsity, indicating broader capacity usage under sparse regimes. Together, these results nuance the common intuition that more superposition leads to more forgetting by showing that overlap interacts with representation strength and capacity allocation. Our toy analysis provides falsifiable hypotheses and diagnostic tools for CL.

Related papers

• How LLMs Follow Instructions: Skillful Coordination, Not a Universal Mechanism [2.5997274006052544]
  We investigate whether instruction-following relies on a universal mechanism or compositional skill deployment.<n>General probes trained across all tasks consistently underperform task-specific specialists, indicating limited representational sharing.<n> temporal analysis shows constraint satisfaction operates as dynamic monitoring during generation rather than pre-generation planning.
  arXiv  Detail & Related papers  (2026-04-07T16:12:52Z)
• Farther the Shift, Sparser the Representation: Analyzing OOD Mechanisms in LLMs [100.02824137397464]
  We investigate how Large Language Models adapt their internal representations when encountering inputs of increasing difficulty.<n>We reveal a consistent and quantifiable phenomenon: as task difficulty increases, the last hidden states of LLMs become substantially sparser.<n>This sparsity--difficulty relation is observable across diverse models and domains.
  arXiv  Detail & Related papers  (2026-03-03T18:48:15Z)
• Temporal Difference Learning with Constrained Initial Representations [41.31941267662611]
  We introduce the Tanh function into the initial layer to fulfill such a constraint.<n>We present our Constrained Initial Representations framework, tagged CIR, which is made up of three components.<n> Empirical results show that CIR exhibits strong performance on numerous continuous control tasks.
  arXiv  Detail & Related papers  (2026-02-12T10:27:57Z)
• State Rank Dynamics in Linear Attention LLMs [37.607046806053035]
  State Rank Stratification is characterized by a distinct spectral bifurcation among linear attention heads.<n>Low-rank heads are indispensable for model reasoning, whereas high-rank heads exhibit significant redundancy.<n>We propose Joint Rank-Norm Pruning, a zero-shot strategy that achieves a 38.9% reduction in KV-cache overhead while largely maintaining model accuracy.
  arXiv  Detail & Related papers  (2026-02-02T15:00:42Z)
• MIND: From Passive Mimicry to Active Reasoning through Capability-Aware Multi-Perspective CoT Distillation [16.96094045628127]
  Existing approaches restrict students to following a single golden rationale and treat different reasoning paths independently.<n>This misalignment leads to a degeneration of the student's latent reasoning distribution, causing suboptimal performance.<n>We propose MIND, a capability-filtered framework that transitions passive mimicry to active cognitive construction.
  arXiv  Detail & Related papers  (2026-01-07T09:08:59Z)
• Emergence of Superposition: Unveiling the Training Dynamics of Chain of Continuous Thought [64.43689151961054]
  We theoretically analyze the training dynamics of a simplified two-layer transformer on the directed graph reachability problem.<n>Our analysis reveals that during training using continuous thought, the index-matching logit will first increase and then remain bounded under mild assumptions.
  arXiv  Detail & Related papers  (2025-09-27T15:23:46Z)
• Learning Action-based Representations Using Invariance [18.1941237781348]
  We introduce action-bisimulation encoding, which learns a multi-step controllability metric that discounts distant state features that are relevant for control. We demonstrate that action-bisimulation pretraining on reward-free, uniformly random data improves sample efficiency in several environments.
  arXiv  Detail & Related papers  (2024-03-25T02:17:54Z)
• Learning Interpretable Policies in Hindsight-Observable POMDPs through Partially Supervised Reinforcement Learning [57.67629402360924]
  We introduce the Partially Supervised Reinforcement Learning (PSRL) framework. At the heart of PSRL is the fusion of both supervised and unsupervised learning. We show that PSRL offers a potent balance, enhancing model interpretability while preserving, and often significantly outperforming, the performance benchmarks set by traditional methods.
  arXiv  Detail & Related papers  (2024-02-14T16:23:23Z)
• Intensity Profile Projection: A Framework for Continuous-Time Representation Learning for Dynamic Networks [50.2033914945157]
  We present a representation learning framework, Intensity Profile Projection, for continuous-time dynamic network data. The framework consists of three stages: estimating pairwise intensity functions, learning a projection which minimises a notion of intensity reconstruction error. Moreoever, we develop estimation theory providing tight control on the error of any estimated trajectory, indicating that the representations could even be used in quite noise-sensitive follow-on analyses.
  arXiv  Detail & Related papers  (2023-06-09T15:38:25Z)
• Causal Triplet: An Open Challenge for Intervention-centric Causal Representation Learning [98.78136504619539]
  Causal Triplet is a causal representation learning benchmark featuring visually more complex scenes. We show that models built with the knowledge of disentangled or object-centric representations significantly outperform their distributed counterparts.
  arXiv  Detail & Related papers  (2023-01-12T17:43:38Z)
• Interpretable Time-series Representation Learning With Multi-Level Disentanglement [56.38489708031278]
  Disentangle Time Series (DTS) is a novel disentanglement enhancement framework for sequential data. DTS generates hierarchical semantic concepts as the interpretable and disentangled representation of time-series. DTS achieves superior performance in downstream applications, with high interpretability of semantic concepts.
  arXiv  Detail & Related papers  (2021-05-17T22:02:24Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.