Neural Network Optimization

FR-Spec: Accelerating Large-Vocabulary Language Models via Frequency-Ranked Speculative Sampling ACL 2025 Function-Space Learning Rates ICML 2025 Fuzzy Speculative Decoding for a Tunable Accuracy-Runtime Tradeoff ACL 2025 Gated Delta Networks: Improving Mamba2 with Delta Rule ICLR 2025 GliaNet: Adaptive Neural Network Structure Learning with Glia-Driven CVPR 2025 Global Convergence of Adjoint-Optimized Neural PDEs JMLR 2025 Global curvature for second-order optimization of neural networks ICML 2025 Good regularity creates large learning rate implicit biases: edge of stability, balancing, and catapult JMLR 2025 Gradient Alignment Improves Test-Time Adaptation for Medical Image Segmentation AAAI 2025 Gradient correlation is a key ingredient to accelerate SGD with momentum ICLR 2025 GradPS: Resolving Futile Neurons in Parameter Sharing Network for Multi-Agent Reinforcement Learning ICML 2025 GReaTer: Gradients Over Reasoning Makes Smaller Language Models Strong Prompt Optimizers ICLR 2025 Grokking at the Edge of Numerical Stability ICLR 2025 Grokking Beyond the Euclidean Norm of Model Parameters ICML 2025 GTA: Supervised-Guided Reinforcement Learning for Text Classification with Large Language Models EMNLP 2025 Gumbel Reranking: Differentiable End-to-End Reranker Optimization ACL 2025 HALL-E: Hierarchical Neural Codec Language Model for Minute-Long Zero-Shot Text-to-Speech Synthesis ICLR 2025 Hallucination Detox: Sensitivity Dropout (SenD) for Large Language Model Training ACL 2025 HFT: Half Fine-Tuning for Large Language Models ACL 2025 HiRA: Parameter-Efficient Hadamard High-Rank Adaptation for Large Language Models ICLR 2025 HoPE: A Novel Positional Encoding Without Long-Term Decay for Enhanced Context Awareness and Extrapolation ACL 2025 HOT: Hadamard-based Optimized Training CVPR 2025 How Does DPO Reduce Toxicity? A Mechanistic Neuron-Level Analysis EMNLP 2025 How Feature Learning Can Improve Neural Scaling Laws ICLR 2025 How Gradient descent balances features: A dynamical analysis for two-layer neural networks ICLR 2025