Entity · technique

Recovery Subspace Dimensionality

techniqueactiverecovery-subspace-dimensionality-bbc803c6·1 events·first seen Jun 12, 2026

Aliases: Recovery Subspace Dimensionality

Co-occurring entities

Split CIFAR-100 The Stable Recovery Manifold: Geometric Principles Governing Recoverability in Continual Learning ResNet-18

More like this (12)

Subspace Projection low-rank subspace projection Unstable Features, Reproducible Subspaces: Understanding Seed Dependence in Sparse Autoencoders Latent World Recovery for Multimodal Learning with Missing Modalities The Stable Recovery Manifold: Geometric Principles Governing Recoverability in Continual Learning Latent World Recovery Any-Dimensional Learning by Sampling target-space recovery profiles Recoverable but Not Stationary: Local Linear Structures in Weights and Activations Rank-Constrained Subspace Learning (RCSL)Sparse Subspace-to-Expert Sharing for Task-Agnostic Continual Learning Do LLM Embedding Spaces Recover Expert Structure?

Recent events (1)

5arXiv · cs.LG·Jun 12, 2026·source ↗

Stable Recovery Manifold hypothesis: catastrophic forgetting as accessibility problem, not information destruction

A new arXiv preprint investigates the geometric structure of recoverability in continual learning using Split CIFAR-100 and a sequentially trained ResNet-18. The authors introduce Recovery Subspace Dimensionality (k_t) and find that recovery dimensionality remains stable across tasks (mean k_t = 8.0) despite substantial representational drift, with principal-angle drift strongly predicting recoverability (r = -0.862). The findings support the Stable Recovery Manifold hypothesis: forgotten knowledge remains compactly decodable, reframing catastrophic forgetting as a manifold-alignment and accessibility problem rather than true information loss.

Evaluation and Benchmarking Split CIFAR-100 Recovery Subspace Dimensionality The Stable Recovery Manifold: Geometric Principles Governing Recoverability in Continual Learning +1 more