1912.02178 , archivePrefix=

representative citing papers

citing papers explorer

Showing 17 of 17 citing papers.

• Grokking: Generalization Beyond Overfitting on Small Algorithmic Datasets cs.LG · 2022-01-06 · unverdicted · none · ref 6

  Neural networks exhibit grokking on small algorithmic datasets, achieving perfect generalization well after overfitting.

• Pointwise Generalization in Deep Neural Networks cs.LG · 2026-05-18 · unverdicted · none · ref 102

  Proposes pointwise Riemannian Dimension from feature eigenvalues to derive tighter, representation-aware generalization bounds for deep networks in the nonlinear regime.

• Characterizing Learning in Deep Neural Networks using Tractable Algorithmic Complexity Analysis cs.LG · 2026-05-15 · unverdicted · none · ref 46

  QuBD extends algorithmic complexity estimation to quantized DNN weights, revealing that complexity decreases during learning, increases with overfitting, follows grokking patterns, and correlates with generalization.

• Policy Optimization in Hybrid Discrete-Continuous Action Spaces via Mixed Gradients cs.LG · 2026-05-14 · unverdicted · none · ref 76

  HPO enables unbiased policy optimization in hybrid action spaces by mixing differentiable simulation gradients with score-function estimates, outperforming PPO as continuous dimensions increase.

• Characterizing Optimizer-Dependent Training Dynamics Through Hessian Eigenvector Displacement and Localization cs.LG · 2026-06-29 · unverdicted · none · ref 22

  Hessian eigenvector displacement and inverse participation ratio metrics show SGD stabilizing leading curvature directions while Adam causes more reorganization and parameter localization in MLP training.

• Quantifying and Optimizing Simplicity via Polynomial Representations cs.AI · 2026-05-28 · unverdicted · none · ref 7

  Polynomial representations yield an effective-degree simplicity metric that predicts generalization across tasks and serves as a differentiable regularizer improving performance in classification and RL.

• Trajectory-Based Difficulty Scoring for Reliable Learning on Tabular Data cs.LG · 2026-05-23 · unverdicted · none · ref 5

  TDS uses per-tree prediction trajectories to derive instance difficulty scores that rank errors better than prior hardness measures and improve active learning, selective prediction, and Mondrian conformal prediction on tabular data.

• Differentiable Optimization Layers for Guaranteed Fairness in Deep Learning cs.LG · 2026-05-16 · unverdicted · none · ref 108

  Introduces a fairness layer for deep learning models that guarantees output parity and an online primal-dual algorithm for aggregate fairness guarantees in streaming predictions with small batch sizes.

• Generalization at the Edge of Stability cs.LG · 2026-04-21 · unverdicted · none · ref 38

  Training at the edge of stability causes neural network optimizers to converge on fractal attractors whose effective dimension, measured via a new sharpness dimension from the Hessian spectrum, bounds generalization error in a way not captured by prior trace or norm measures.

• Robust Policy Optimization to Prevent Catastrophic Forgetting cs.LG · 2026-02-09 · unverdicted · none · ref 27

  FRPO applies a max-min robust optimization over KL-bounded policy neighborhoods during RLHF to reduce catastrophic forgetting of safety and accuracy under subsequent SFT or RL fine-tuning.

• Sharpness-Aware Minimization for Efficiently Improving Generalization cs.LG · 2020-10-03 · conditional · none · ref 26

  SAM solves a min-max problem to locate flat low-loss regions, improving generalization on CIFAR, ImageNet and label-noise tasks.

• Flatness Preserves Instruction Following in Vision-Language-Action Models cs.RO · 2026-06-22 · unverdicted · none · ref 29

  Sharpness-aware minimization during VLA finetuning preserves instruction following and yields over 60% gains across simulation and real-world tasks.

• SingGuard: A Policy-Adaptive Multimodal LLM Guardrail with Dynamic Reasoning cs.CV · 2026-06-22 · unverdicted · none · ref 65 · 2 links

  SingGuard introduces a policy-adaptive multimodal LLM guardrail with dynamic reasoning regimes and SingGuard-Bench, reporting SOTA F1 scores across 35 datasets and improved policy-following accuracy under runtime shifts.

• Flatness and Gradient Alignment Are Both Necessary: Spectral-Aware Gradient-Aligned Exploration for Multi-Distribution Learning cs.LG · 2026-05-08 · conditional · none · ref 6

  SAGE, an optimizer combining spectral polar-factor perturbation with gradient-agreement-scaled noise, reports 78.9% average on DomainBed.

• MER-DG: Modality-Entropy Regularization for Multimodal Domain Generalization cs.LG · 2026-05-03 · unverdicted · none · ref 17

  MER-DG applies modality-entropy regularization to reduce fusion overfitting in multimodal domain generalization, reporting average gains of 5% over standard fusion and 2% over prior methods on EPIC-Kitchens and HAC benchmarks.

• An Information-Theoretic Analysis of OOD Generalization in Meta-Reinforcement Learning cs.LG · 2025-10-27 · unverdicted · none · ref 6

  The work establishes OOD generalization bounds for meta-supervised learning and meta-RL that exploit MDP structure, then analyzes a gradient-based meta-RL algorithm.

• TopoGeoScore: A Self-Supervised Source-Only Geometric Framework for OOD Checkpoint Selection cs.LG · 2026-05-09 · unverdicted · none · ref 14

  TopoGeoScore learns a non-negative linear combination of geometric and topological features from source embeddings via self-supervised invariance to select robust checkpoints for OOD scenarios.