regularization

This paper introduces Rotation-Preserving Supervised Fine-Tuning (RPSFT), a method that improves out-of-domain generalization by preserving projected rotations in pretrained singular subspaces during fine-tuning.

This paper introduces FragileFlow, a plug-in regularizer that improves the robustness of LLMs and VLMs by controlling 'correct-but-fragile' predictions through spectral analysis and PAC-Bayes bounds.

This paper introduces DOSER, a framework using diffusion models for out-of-distribution detection and selective regularization in offline reinforcement learning. It aims to improve performance on static datasets by distinguishing between beneficial and detrimental OOD actions.

The authors introduce Sub-JEPA, a method using Subspace Gaussian Regularization to improve the stability of end-to-end world models like LeWM, showing consistent performance gains on continuous-control benchmarks.

The paper introduces OrthoReg, a plug-and-play regularizer that enforces weight orthogonality during fine-tuning to improve task arithmetic and model merging without extra compute.

This paper presents adversarial and virtual adversarial training methods adapted for text classification by applying perturbations to word embeddings in RNNs rather than raw inputs. The approach achieves state-of-the-art results on semi-supervised and supervised text classification benchmarks while reducing overfitting.