{"total":51,"rows":[{"id":"2501.17116","title":"Optimizing Large Language Model Training Using FP4 Quantization","created":"2025-01-28","cited_by_count":56,"area":null},{"id":"2505.19115","title":"FP4 All the Way: Fully Quantized Training of LLMs","created":"2025-05-25","cited_by_count":40,"area":null},{"id":"2505.14669","title":"Quartet: Native FP4 Training Can Be Optimal for Large Language Models","created":"2025-05-20","cited_by_count":39,"area":null},{"id":"2502.02732","title":"Peri-LN: Revisiting Normalization Layer in the Transformer Architecture","created":"2025-02-04","cited_by_count":34,"area":null},{"id":"2410.19313","title":"COAT: Compressing Optimizer states and Activation for Memory-Efficient FP8 Training","created":"2024-10-25","cited_by_count":33,"area":null},{"id":"2504.20966","title":"Softpick: No Attention Sink, No Massive Activations with Rectified Softmax","created":"2025-04-29","cited_by_count":22,"area":null},{"id":"2506.08027","title":"Recipes for Pre-training LLMs with MXFP8","created":"2025-05-30","cited_by_count":22,"area":null},{"id":"2502.18934","title":"Kanana: Compute-efficient Bilingual Language Models","created":"2025-02-26","cited_by_count":20,"area":null},{"id":"2411.09909","title":"AMXFP4: Taming Activation Outliers with Asymmetric Microscaling Floating-Point for 4-bit LLM Inference","created":"2024-11-15","cited_by_count":19,"area":null},{"id":"2505.14302","title":"Scaling Law for Quantization-Aware Training","created":"2025-05-20","cited_by_count":17,"area":null},{"id":"2501.02625","title":"HALO: Hadamard-Assisted Lower-Precision Optimization for LLMs","created":"2025-01-05","cited_by_count":15,"area":null},{"id":"2503.08040","title":"Accurate INT8 Training Through Dynamic Block-Level Fallback","created":"2025-03-11","cited_by_count":14,"area":null},{"id":"2505.10949","title":"FP64 is All You Need: Rethinking Failure Modes in Physics-Informed Neural Networks","created":"2025-05-16","cited_by_count":12,"area":null},{"id":"2505.01043","title":"Low-Precision Training of Large Language Models: Methods, Challenges, and Opportunities","created":"2025-05-02","cited_by_count":11,"area":null},{"id":"2505.20524","title":"Towards Fully FP8 GEMM LLM Training at Scale","created":"2025-05-26","cited_by_count":11,"area":null},{"id":"2603.00040","title":"Attn-QAT: 4-Bit Attention With Quantization-Aware Training","created":"2026-02-09","cited_by_count":8,"area":null},{"id":"2607.04422","title":"Full-Stack FP4: Stable LLM Pretraining with Quantized Projections, Optimizers, and Attention","created":"2026-07-05","cited_by_count":8,"area":null},{"id":"2602.22732","title":"Generative Recommendation for Large-Scale Advertising","created":"2026-02-26","cited_by_count":7,"area":null},{"id":"2511.23225","title":"TWEO: Transformers Without Extreme Outliers Enables FP8 Training And Quantization For Dummies","created":"2025-11-28","cited_by_count":6,"area":null},{"id":"2603.16731","title":"Understanding Quantization of Optimizer States in LLM Pre-training: Dynamics of State Staleness and Effectiveness of State Resets","created":"2026-03-17","cited_by_count":6,"area":null},{"id":"2411.13010","title":"Deriving Activation Functions Using Integration","created":"2024-11-20","cited_by_count":5,"area":null},{"id":"2502.05967","title":"$\\mu$nit Scaling: Simple and Scalable FP8 LLM Training","created":"2025-02-09","cited_by_count":5,"area":null},{"id":"2510.09904","title":"Stability of Transformers under Layer Normalization","created":"2025-10-10","cited_by_count":5,"area":null},{"id":"2602.23349","title":"FlashOptim: Optimizers for Memory-Efficient Training","created":"2026-02-26","cited_by_count":5,"area":null},{"id":"2604.15416","title":"StoSignSGD: Unbiased Structural Stochasticity Fixes SignSGD for Training Large Language Models","created":"2026-04-16","cited_by_count":5,"area":null},{"id":"2506.20752","title":"Characterization and Mitigation of Training Instabilities in Microscaling Formats","created":"2025-06-25","cited_by_count":4,"area":null},{"id":"2509.17791","title":"Elucidating the Design Space of FP4 training","created":"2025-09-22","cited_by_count":4,"area":null},{"id":"2602.08923","title":"DynamiQ: Accelerating Gradient Synchronization using Compressed Multi-hop All-reduce","created":"2026-02-09","cited_by_count":4,"area":null},{"id":"2602.02047","title":"Dissecting Outlier Dynamics in LLM NVFP4 Pretraining","created":"2026-02-02","cited_by_count":3,"area":null},{"id":"2503.09975","title":"Faster Inference of LLMs using FP8 on the Intel Gaudi","created":"2025-03-13","cited_by_count":2,"area":null},{"id":"2510.04202","title":"Spectral Alignment as Predictor of Loss Explosion in Neural Network Training","created":"2025-10-05","cited_by_count":2,"area":null},{"id":"2511.05811","title":"MOSS: Efficient and Accurate FP8 LLM Training with Microscaling and Automatic Scaling","created":"2025-11-08","cited_by_count":2,"area":null},{"id":"2512.15306","title":"LLMQ: Efficient Lower-Precision Pretraining for Consumer GPUs","created":"2025-12-17","cited_by_count":2,"area":null},{"id":"2602.02400","title":"An Empirical Study on Noisy Data and LLM Pretraining Loss Divergence","created":"2026-02-02","cited_by_count":2,"area":null},{"id":"2510.24061","title":"FALQON: Accelerating LoRA Fine-tuning with Low-Bit Floating-Point Arithmetic","created":"2025-10-28","cited_by_count":1,"area":null},{"id":"2605.10886","title":"LoKA: Low-precision Kernel Applications for Recommendation Models At Scale","created":"2026-05-11","cited_by_count":1,"area":null},{"id":"2605.20402","title":"Decomposing MXFP4 quantization error for LLM reinforcement learning: reducible bias, recoverable deadzone, and an irreducible floor","created":"2026-05-19","cited_by_count":1,"area":null},{"id":"2605.23191","title":"Expand More, Shrink Less: Shaping Effective-Rank Dynamics for Dense Scaling in Recommendation","created":"2026-05-22","cited_by_count":1,"area":null},{"id":"2607.10611","title":"M+Adam: Low-Precision Training via Additive-Multiplicative Optimization","created":"2026-07-12","cited_by_count":1,"area":null},{"id":"2505.11170","title":"Gaussian Weight Sampling for Scalable, Efficient and Stable Pseudo-Quantization Training","created":"2025-05-16","cited_by_count":0,"area":null},{"id":"2512.04051","title":"Convergence for Discrete Parameter Update Schemes","created":"2025-12-03","cited_by_count":0,"area":null},{"id":"2512.22804","title":"MoR: Mixture Of Representations For Mixed-Precision Training","created":"2025-12-28","cited_by_count":0,"area":null},{"id":"2602.20646","title":"On the Convergence of Stochastic Gradient Descent with Perturbed Forward-Backward Passes","created":"2026-02-24","cited_by_count":0,"area":null},{"id":"2605.09370","title":"From Detection to Recovery: Operational Analysis on LLM Pre-training with 504 GPUs","created":"2026-05-10","cited_by_count":0,"area":null},{"id":"2608.02091","title":"One QK Channel, Many Sources: Tracing Low-Precision Attention Collapse","created":"2026-08-03","cited_by_count":0,"area":null},{"id":"2608.07323","title":"Is SwiGLU's Open Positive Tail Necessary? Evidence from Closed-Tail Gating with MemGLU","created":"2026-08-07","cited_by_count":0,"area":null},{"id":"2608.29667","title":"A Target-Centric Survey of Quantization-Aware Training","created":"2026-08-30","cited_by_count":0,"area":null},{"id":"2609.34272","title":"Broken Symmetry in BF16 Attention: Why FlashAttention Gradients Blow Up Late in Training","created":"2026-09-28","cited_by_count":0,"area":null},{"id":"2609.35392","title":"The Hidden Ratio in Adam: Stable Structure, Compression, and Sign Dynamics","created":"2026-09-28","cited_by_count":0,"area":null},{"id":"2609.39813","title":"Backward-State Policy Is Part of the Learning Algorithm","created":"2026-09-30","cited_by_count":0,"area":null},{"id":"2610.17637","title":"Mixed Precision Scaling Laws for Quantized Pretraining","created":"2026-10-13","cited_by_count":0,"area":null}],"next_cursor":null}