Delta-Matching: Closing the Final Gap of Native 8-bit Training for LLMs
cs.LG
Submitted: 2026-09-29
Updated: 2026-09-29
Project page: https://nvidia.github.io/cudnn-frontend/mxfp8-attentionscaling
Sources
- Language Models Enable Simple Systems for Generating Structured Views of Heterogeneous Data Lakes
- Lessons from the Trenches on Reproducible Evaluation of Language Models
- Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge
- DeepSeek-V3 Technical Report
- Full-Stack FP4: Stable LLM Pretraining with Quantized Projections, Optimizers, and Attention
- Mamba: Linear-Time Sequence Modeling with Selective State Spaces
- RULER: What's the Real Context Size of Your Long-Context Language Models?
- Mechanism-Driven Monitors for Preemptive Detection of LLM Training Instability
- Kimi Linear: An Expressive, Efficient Attention Architecture
- Kimi K3: Open Frontier Intelligence
- Kimi K2: Open Agentic Intelligence
- Adam: A Method for Stochastic Optimization
- Every Activation Boosted: Scaling General Reasoner to 1 Trillion Open Language Foundation
- Muon is Scalable for LLM Training
- Decoupled Weight Decay Regularization
- Mixed Precision Training
- FP8 Formats for Deep Learning
- MiniMax-01: Scaling Foundation Models with Lightning Attention
- Recipes for Pre-training LLMs with MXFP8
- Nemotron-H: A Family of Accurate and Efficient Hybrid Mamba-Transformer 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