Same Methods, Different Rankings: Trainable Depth as an Evaluation Variable in Continual Learning

arXiv:2604.21927 · cs.LG · Submitted 2026-04-23 · Read on arXiv

cs.LG

Submitted: 2026-04-23

Updated: 2026-10-05

Comments: 14 pages, 4 figures

Journal ref: P.-T. Iordache, E. Burceanu and M. Dascalu, "Same Methods, Different Rankings: Trainable Depth as an Evaluation Variable in Continual Learning," IEEE Access, vol. 14, pp. 136009-136022, 2026

DOI: 10.1109/ACCESS.2026.3728110

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• Fine-Tuning — 3× in this paper · explained in 5 episodes
• Adaptation — 2× in this paper · explained in 1 episode
• Continual Learning — 2× in this paper · explained in 2 episodes
• Benchmark — 1× in this paper · explained in 2 episodes
• Benchmarking — 1× in this paper · explained in 4 episodes

Sources

• Continual Learning Through Synaptic Intelligence
• Learning without Forgetting
• Gradient Episodic Memory for Continual Learning
• Towards Robust Evaluations of Continual Learning
• Three scenarios for continual learning
• An Empirical Investigation of Catastrophic Forgetting in Gradient-Based Neural Networks
• Progress & Compress: A scalable framework for continual learning
• Fine-Tuning can Distort Pretrained Features and Underperform Out-of-Distribution
• Continual Lifelong Learning in Natural Language Processing: A Survey
• Re-evaluating Continual Learning Scenarios: A Categorization and Case for Strong Baselines
• Continual World: A Robotic Benchmark For Continual Reinforcement Learning
• Continual Learning in Linear Classification on Separable Data
• The Effect of Task Ordering in Continual Learning
• Scalable and Order-robust Continual Learning with Additive Parameter Decomposition
• Subspace Geometry Governs Catastrophic Forgetting in Low-Rank Adaptation
• Transfusion: Understanding Transfer Learning for Medical Imaging
• Critical Learning Periods in Deep Neural Networks
• Analyzing and Reducing Catastrophic Forgetting in Parameter Efficient Tuning
• Budgeted Online Continual Learning by Adaptive Layer Freezing and Frequency-based Sampling
• Hyperparameters in Continual Learning: A Reality Check

Related papers

• Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
• AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
• Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
• Convergence issues in Relational Concept Analysis based on AOC-posets
• Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
• Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks