Beyond Endpoint Scores: Time- and Capacity-Conditioned Evaluation of Continual Knowledge Updating
cs.LG
Submitted: 2026-09-03
Updated: 2026-09-03
Comments: 13 pages, 4 figures. Extended preprint
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- Understanding LoRA as Knowledge Memory: An Empirical Analysis
- Continual evaluation for lifelong learning: Identifying the stability gap
- ForgetBench: Benchmarking Forgetting Dynamics of Long-Term Parametric Memory in Language Models
- When Does Continual Learning Require Learning
- Aging with GRACE: Lifelong Model Editing with Discrete Key-Value Adaptors
- Same Methods, Different Rankings: Trainable Depth as an Evaluation Variable in Continual Learning
- TemporalWiki: A Lifelong Benchmark for Training and Evaluating Ever-Evolving Language Models
- MoRA: High-Rank Updating for Parameter-Efficient Fine-Tuning
- Can Large Language Models Keep Up? Benchmarking Online Adaptation to Continual Knowledge Streams
- Continual Fine-Tuning of Large Language Models via Program Memory
- Energy-Structured Low-Rank Adaptation for Continual Learning
- RAG or Learning? Understanding the Limits of LLM Adaptation under Continuous Knowledge Drift in the Real World
- Edit Less, Achieve More: Dynamic Sparse Neuron Masking for Lifelong Knowledge Editing in LLMs
- Take Only What You Need: Rank Minimization as an Implicit Forgetting Regularizer in Continual Learning
- EvoEdit: Evolving Null-space Alignment for Robust and Efficient Knowledge Editing
- Can a Language Model Learn Facts Continually in Its Weights?
- Continual Knowledge Updating in LLM Systems: Learning Through Multi-Timescale Memory Dynamics
- Merge before Forget: A Single LoRA Continual Learning via Continual Merging
- MEMOIR: Lifelong Model Editing with Minimal Overwrite and Informed Retention for LLMs
- WISE: Rethinking the Knowledge Memory for Lifelong Model Editing of Large Language Models
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