| Challenge: | Muon’s orthogonalized update rule suppresses the emergence of heavy-tailed weight spectra and over-emphasizes the training along noise-dominated directions. |
| Approach: | They propose to preserve Muon's ability to capture parameter interdependencies while producing heavier-tailed updates and inducing heavier-tail weight spectra. |
| Outcome: | The proposed algorithm suppresses the emergence of heavy-tailed weight spectra and over-emphasizes training along noise-dominated directions. |
Similar Papers
LLM Braces: Straightening Out LLM Predictions with Relevant Sub-Updates (2025.acl-long)
Copied to clipboard
| Challenge: | Recent studies reveal that much of the knowledge in a Transformer-based Large Language Model (LLM) is encoded in its feed-forward (FFN) layers, where each FNN layer can be interpreted as the summation of sub-updates, each corresponding to a weighted column vector from the FFN’s value parameter matrix. |
| Approach: | They propose a method that computes relevance scores associated with value vectors in FFN layers and leverages these scores to dynamically adjust the contribution of sub-updates. |
| Outcome: | The proposed framework outperforms baseline approaches in fine-tuning and zero-shot settings while requiring significantly fewer tunable parameters. |
Model Balancing Helps Low-data Training and Fine-tuning (2024.emnlp-main)
Copied to clipboard
| Challenge: | Recent advances in foundation models have emphasized the need to align pre-trained models with specialized domains using small, curated datasets. |
| Approach: | They propose a layer-wise learning rate scheduler that balances training quality across layers . they adapt it to a curated dataset to achieve alignment with specialized domains . |
| Outcome: | The proposed model shows that it can be used to balance training quality across layers and improve low-data training and fine-tuning for both NLP and SciML tasks. |
Outlier-Safe Pre-Training for Robust 4-Bit Quantization of Large Language Models (2025.acl-long)
Copied to clipboard
| Challenge: | Quantization is a practical solution for deploying Large Language Models in resource-constrained environments. |
| Approach: | They propose an outlier-safe pre-training approach that prevents outlier formation . they validate a 1.4B-parameter model on 1 trillion tokens with no outliers . |
| Outcome: | The proposed model achieves a 35.7 average score on 1 trillion tokens with 2% training overhead. |
AdamS: Momentum Itself Can Be A Normalizer for LLM Pretraining and Post-training (2025.emnlp-main)
Copied to clipboard
| Challenge: | Empirically, AdamS demonstrates strong performance in various tasks . et al., 2023b): AdamS is efficient, efficient, and model-agnostic. |
| Approach: | They propose a model-agnostic alternative to Adam for large language model pretraining and post-training. |
| Outcome: | The proposed method matches memory footprint of SGD with momentum while delivering superior performance. |
BWLA: Breaking the Barrier of W1AX Post-Training Quantization for LLMs (2026.acl-long)
Copied to clipboard
| Challenge: | Large language models have driven major progress in NLP, but memory and compute requirements hinder practical deployment. |
| Approach: | They propose a framework that preserves high accuracy while achieving 1-bit weight quantization . the orthogonal-kronecker transformation learns an orthogonale mapping via EM minimization - a new approach to quantization is proposed . |
| Outcome: | The proposed framework achieves 1-bit weight quantization with low activations with low-bit activations. |
The Lower The Simpler: Simplifying Hierarchical Recurrent Models (N19-1)
Copied to clipboard
| Challenge: | Using a simplified version of GRU, we replace the GRUs at the middle layers of hierarchical recurrent models with Fixed-size Ordinally-Forgetting Encoding (FOFE). |
| Approach: | They propose to make the lower layers simpler than the upper ones to simplify two typical hierarchical recurrent models, namely Hierarchical Recurrent Encoder-Decoder (HRED) and R-NET, whose basic building block is GRU. |
| Outcome: | The proposed models contain less trainable parameters, consume less training time, and achieve slightly better performance than baseline models. |
BalanceSFT: Improving LLM Function Calling with Balanced Training Signals and Data Hardness (2026.findings-acl)
Copied to clipboard
Bingguang Hao, Zengzhuang Xu, Maolin Wang, Yuntao Wen, Yicheng Chen, Cunyin Peng, Long Chen, Xiangyu Zhao, Jinjie Gu, Chenyi Zhuang, Ji Zhang
| Challenge: | Currently, Supervised Fine-Tuning (SFT) is the prevailing method for equipping Large Language Models (LLMs) with function calling capabilities, but its effectiveness is often compromised by two challenges: 1) lengthy Chain-of-Thought (CoT) reasoning tokens dominate training signals over concise function calls in the learning objective; 2) scarcity of hard training examples. |
| Approach: | They propose a framework that uses a self-adjusted signal balancing loss and a hard data re-sampling strategy to selectively generate new, high-quality complex data guided by model errors. |
| Outcome: | The proposed framework surpasses state-of-the-art models like GPT-5 in function calling performance. |
WISCA: A Lightweight Model Transition Method to Improve LLM Training via Weight Scaling (2026.findings-acl)
Copied to clipboard
Jiacheng Li, Jianchao Tan, Zhidong Yang, Pingwei Sun, Feiye Huo, Jiayu Qin, Xiangyu Zhang, Maoxin He, Guangming Tan, Weile Jia, Xunliang Cai, Tong Zhao
| Challenge: | Recent advances in training optimization for Transformer-based large language models lack systematic optimization of weight patterns during training. |
| Approach: | They propose a Weight Scaling method that rescales weights while preserving model outputs to improve model training efficiency and model quality. |
| Outcome: | The proposed method significantly improves convergence quality and loss reduction in LLMs with Grouped Query Attention architectures and LoRA fine-tuning tasks. |
Data Weighted Training Strategies for Grammatical Error Correction (2020.tacl-1)
Copied to clipboard
| Challenge: | Recent advances in the task of Grammatical Error Correction (GEC) have been driven by addressing data sparsity, both through new methods for generating large and noisy pretraining data and through the publication of small and higher-quality finetuning data in the BEA-2019 shared task. |
| Approach: | They propose to incorporate delta-log-perplexity, a type of example scoring, into a training schedule for Grammatical Error Correction (GEC) they perform experiments that shed light on the function and applicability of delta- log-perplicity. |
| Outcome: | The proposed methods incorporate delta-log-perplexity, a type of example scoring, into a training schedule for the task. |
Hybrid-Regressive Paradigm for Accurate and Speed-Robust Neural Machine Translation (2023.findings-acl)
Copied to clipboard
| Challenge: | Autoregressive translation (NAT) is less robust in decoding batch size and hardware settings than NAT. |
| Approach: | They propose a two-stage translation prototype that prompts a small number of AT predictions and fills in previously skipped tokens at once. |
| Outcome: | The proposed translation prototype achieves comparable translation quality with AT while having 1.5x faster inference speed regardless of batch size and device. |