Half-S: Halving the Scale for Near-Lossless 4-Bit LLM Training (2026.findings-acl)
Copied to clipboard
Jinyang Du, Ruihao Gong, Linghan Ai, Zining Wang, Yunke Peng, Yao Wang, Lei Yan, null Wxuefei, Yaoyuan Wang, Jinyang Guo, Dahua Lin, Xianglong Liu
| Challenge: | Existing 4-bit training pipelines rely on max-scaling, which causes representation collapse . despite this, there are limitations in the accuracy of 4-bit LLM training . |
| Approach: | They propose a scaling strategy that uses half-scaling as a hardware-friendly default . they propose fp4 support that allows for a faster scaling of large language models . |
| Outcome: | The proposed scaling strategy narrows the gap between theoretical optimum and BF16 while maintaining the efficiency benefits of 4-bit training. |
Similar Papers
LLM-FP4: 4-Bit Floating-Point Quantized Transformers (2023.emnlp-main)
Copied to clipboard
| Challenge: | Existing quantization solutions are integer-based and struggle with bit widths below 8 bits. |
| Approach: | They propose a method for quantizing weights and activations in large language models down to 4-bit floating-point values in a post-training manner. |
| Outcome: | The proposed method outperforms existing methods on common sense zero-shot reasoning tasks by 12.7 points. |
Achieving binary weight and activation for LLMs using Post-Training Quantization (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing methods for quantizing large language models suffer from performance degradation when weights are quantized to 1 bit. |
| Approach: | They propose a post-training quantization framework with W(1+1)A(14) configuration . they propose utilizing Hessian-aware fine-grained grouping along with an EM-based quantization scheme . |
| Outcome: | The proposed method surpasses state-of-the-art (SOTA) LLM quantization baselines on W2A4 across multiple tasks. |
Revisiting Block-based Quantisation: What is Important for Sub-8-bit LLM Inference? (2023.emnlp-main)
Copied to clipboard
| Challenge: | Existing quantisation methods mainly focus on 8-bit LLMs . a lack of scaling offsets in the quantisation process limits the use of LLM inference. |
| Approach: | They propose to use block quantisations to reduce scaling offsets in Large language models . they find that the block quantizations reduce scaling only from an arithmetic perspective . |
| Outcome: | The proposed methods reduce scaling offsets solely from an arithmetic perspective without additional treatments in the computational path. |
HFT: Half Fine-Tuning for Large Language Models (2025.acl-long)
Copied to clipboard
| Challenge: | Large language models (LLMs) with one or more fine-tuning phases can unlock various capabilities, but can be catastrophic forgetting during sequential training. |
| Approach: | They propose a method to regularly reset partial parameters to mitigate forgetting issues by using half fine-tuning instead of full fine-uning. |
| Outcome: | The proposed approach reduces the risk of catastrophic forgetting during training and the parametric knowledge lost during training may be overwhelmed by incoming training data. |
LM-Infinite: Zero-Shot Extreme Length Generalization for Large Language Models (2024.naacl-long)
Copied to clipboard
| Challenge: | Currently, large language models (LLMs) train on short text segments due to the computational overhead quadratic in the input lengths of their Transformer architectures. |
| Approach: | They propose a method that allows LLMs pre-trained with 2K or 4K-long segments to generalize to up to 200M length inputs while retaining perplexity. |
| Outcome: | The proposed method achieves 2.7 decoding speed up and 7.5 memory saving over the original model. |
PTQ1.61: Push the Real Limit of Extremely Low-Bit Post-Training Quantization Methods for Large Language Models (2025.acl-long)
Copied to clipboard
| Challenge: | Existing methods for sub 2-bit quantization introduce an extra 1-bit or more per weight. |
| Approach: | They propose a sub 2-bit post-training quantization method that enables weight quantization to 1.61-bit for the first time. |
| Outcome: | The proposed method reduces the upper bound of quantization error to 1.61-bit for the first time. |
Low-Bit Quantization Favors Undertrained LLMs (2025.acl-long)
Copied to clipboard
| Challenge: | Larger models or those trained on fewer tokens exhibit less quantization-induced degradation (QiD), while smaller, well-trained models face significant performance losses. |
| Approach: | They propose to use QiD to measure an LLM’s training levels and determine the number of training tokens required for fully training LLMs of various sizes. |
| Outcome: | The proposed scaling laws can predict the quantization performance of different-sized LLMs trained with tokens. |
Enhancing Computation Efficiency in Large Language Models through Weight and Activation Quantization (2023.emnlp-main)
Copied to clipboard
| Challenge: | Large language models (LLMs) are proficient in natural language processing tasks, but their deployment is limited by extensive parameter sizes and computational demands. |
| Approach: | They propose a method to enhance computational efficiency in large language models by 4-bit weight and 8-bit activation quantization. |
| Outcome: | The proposed techniques significantly boost task accuracies to levels comparable with full-precision models. |
Bit-by-Bit: Progressive QAT Strategy with Outlier Channel Splitting for Stable Low-Bit LLMs (2026.acl-long)
Copied to clipboard
Binxing Xu, Hao Gu, Lujun Li, Hao Wang, Bei Liu, Jiacheng Liu, Qiyuan Zhu, Xintong Yang, Chao Li, Sirui Han, Yike Guo
| Challenge: | Existing approaches to training LLMs at ultra-low precisions suffer from convergence instability and substantial training costs. |
| Approach: | They propose a progressive QAT framework with outlier channel splitting to address these issues . they use nested structure of integer quantization grids to enable a "train once, deploy any precision" paradigm . |
| Outcome: | The proposed framework outperforms baselines on both Llama2/3 and W2A16, with an 11 speedup over BF16. |
Scaling Down, Serving Fast: Compressing and Deploying Efficient LLMs for Recommendation Systems (2025.emnlp-industry)
Copied to clipboard
Kayhan Behdin, Ata Fatahibaarzi, Qingquan Song, Yun Dai, Aman Gupta, Zhipeng Wang, Hejian Sang, Shao Tang, Gregory Dexter, Sirou Zhu, Siyu Zhu, Tejas Dharamsi, Vignesh Kothapalli, Zhoutong Fu, Yihan Cao, Pin-Lun Hsu, Fedor Borisyuk, Natesh S. Pillai, Luke Simon, Rahul Mazumder
| Challenge: | Large language models (LLMs) have demonstrated remarkable performance across a wide range of industrial applications. |
| Approach: | They propose two techniques for training and deploying small language models that deliver high performance for a variety of industry use cases. |
| Outcome: | The proposed techniques retain much of the quality of larger models while reducing training/serving costs and latency. |