The Intruder Threshold: A Spectral Law for LoRA Fine-Tuning
cs.LG, stat.ML
Submitted: 2026-07-26
Updated: 2026-09-12
Code: https://github.com/huggingface/peft
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- LoRA vs Full Fine-tuning: An Illusion of Equivalence
- Understanding Catastrophic Forgetting In LoRA via Mean-Field Attention Dynamics
- GAIN: Multiplicative Modulation for Domain Adaptation
- Spectral phase transitions and trainability in neural network learning dynamics
- Random Matrix Theory of Early-Stopped Gradient Flow: A Transient BBP Scenario
- Discontinuous BBP transitions
- High-Dimensional Theory of LoRA Fine-Tuning in a Solvable Attention Model
- Spectral Dynamics in Deep Networks: Feature Learning, Outlier Escape, and Learning Rate Transfer
- OPLoRA: Orthogonal Projection LoRA Prevents Catastrophic Forgetting during Parameter-Efficient Fine-Tuning
- A Rank Stabilization Scaling Factor for Fine-Tuning with LoRA
- Spectral Unforgetting: Post-Hoc Recovery of Damaged Capabilities Without Retraining
- VeFA: Vector-Based Feature Space Adaptation for Robust Model Fine-Tuning
- An Empirical Study of Catastrophic Forgetting in Large Language Models During Continual Fine-tuning
- Scaling Laws for Forgetting When Fine-Tuning Large Language Models
- Qwen2.5 Technical Report
- The Llama 3 Herd of Models
- Mistral 7B
- 2 OLMo 2 Furious
- Mamba: Linear-Time Sequence Modeling with Selective State Spaces
- OLMoE: Open Mixture-of-Experts 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