GSQ Collection GSQ: Highly-Accurate Low-Precision Scalar Quantization for LLMs via Gumbel-Softmax Sampling, https://huggingface.co/papers/2604.18556 • 9 items • Updated 6 days ago • 5
GSQ: Highly-Accurate Low-Precision Scalar Quantization for LLMs via Gumbel-Softmax Sampling Paper • 2604.18556 • Published Apr 20 • 7
Quartet: Native FP4 Training Can Be Optimal for Large Language Models Paper • 2505.14669 • Published May 20, 2025 • 78
HALO: Hadamard-Assisted Lossless Optimization for Efficient Low-Precision LLM Training and Fine-Tuning Paper • 2501.02625 • Published Jan 5, 2025 • 15
QuEST: Stable Training of LLMs with 1-Bit Weights and Activations Paper • 2502.05003 • Published Feb 7, 2025 • 44