Neural Network Optimization

Pioneering 4-Bit FP Quantization for Diffusion Models: Mixup-Sign Quantization and Timestep-Aware Fine-Tuning CVPR 2025 Preserving Deep Representations in One-Shot Pruning: A Hessian-Free Second-Order Optimization Framework ICLR 2025 QuEST: Low-bit Diffusion Model Quantization via Efficient Selective Finetuning ICCV 2025 Rethinking Neural-based Matrix Inversion: Why can’t, and Where can AISTATS 2025 Revealing the Deceptiveness of Knowledge Editing: A Mechanistic Analysis of Superficial Editing ACL 2025 Run LoRA Run: Faster and Lighter LoRA Implementations ACL 2025 Scheduling Weight Transitions for Quantization-Aware Training ICCV 2025 SEED: Accelerating Reasoning Tree Construction via Scheduled Speculative Decoding COLING 2025 Seeking Consistent Flat Minima for Better Domain Generalization via Refining Loss Landscapes CVPR 2025 Segment-Based Attention Masking for GPTs ACL 2025 Self-Attentive Spatio-Temporal Calibration for Precise Intermediate Layer Matching in ANN-to-SNN Distillation AAAI 2025 Sequence-level Large Language Model Training with Contrastive Preference Optimization NAACL 2025 Sketch-of-Thought: Efficient LLM Reasoning with Adaptive Cognitive-Inspired Sketching EMNLP 2025 SOAP: Improving and Stabilizing Shampoo using Adam for Language Modeling ICLR 2025 Solving Token Gradient Conflict in Mixture-of-Experts for Large Vision-Language Model ICLR 2025 Spectral Scaling Laws in Language Models: emphHow Effectively Do Feed-Forward Networks Use Their Latent Space? EMNLP 2025 Stabilizing Sharpness-Aware Minimization Through A Simple Renormalization Strategy JMLR 2025 Star with Bilinear Mapping CVPR 2025 Stepwise Reasoning Checkpoint Analysis: A Test Time Scaling Method to Enhance LLMs’ Reasoning EMNLP 2025 Stochastic Rounding for LLM Training: Theory and Practice AISTATS 2025 Straight to Zero: Why Linearly Decaying the Learning Rate to Zero Works Best for LLMs ICLR 2025 SuBiTO: Synopsis-based Training Optimization for Continuous Real-Time Neural Learning over Big Streaming Data AAAI 2025 Subnet-Aware Dynamic Supernet Training for Neural Architecture Search CVPR 2025 Taming Transformer Without Using Learning Rate Warmup ICLR 2025 The AdEMAMix Optimizer: Better, Faster, Older ICLR 2025