Large language models (LLMs) sometimes exhibit language confusion when generating non-English text. Existing approaches typically rely on fine-tuning to mitigate this issue. In contrast, we propose a tuning-free paradigm for reducing language confusion. Within this paradigm, we introduce two methods: Language-Aware Token Boosting (LATB), which applies targeted perturbations to tokens associated with the desired language, and Adaptive Language-Aware Token Boosting (Adaptive-LATB), which dynamically adjusts these perturbations based on the model's confidence in the intended language. Experiments demonstrate that our methods effectively improve multilingual alignment by reducing language confusion, while maintain the summarization quality without requiring any additional fine-tuning. Our code is publicly available. https://github.com/scbdatax/genai-datax-language-aware-token-boosting.
Better Pre-Training by Reducing Representation Confusion
Tencent（腾讯）；Peking University（北京大学）
在这项工作中，我们重新审视了基于transformer的预训练语言模型，并分别在位置编码和模型表示中识别出两种不同类型的信息混淆。首先，我们表明，在相对位置编码，相对距离和方向的联合建模带来了两个异构信息之间的混淆。这可能会使模型无法捕获相同距离和相反方向的关联语义，进而影响下游任务的性能。其次，我们注意到带有MLM（Mask Language Modeling）预训练目标的BERT输出相似的token表示（不同token的最后隐藏状态）和head表示（不同head的注意力权重），这可能使不同token和head所表达的信息的多样性受到限制。基于上述研究，我们提出了两种改进预训练语言模型的新技术：解耦方向相对位置（DDRP）编码和MTH预训练目标。DDRP是对经典相对位置编码中的相对距离特征和方向特征的继承。MTH除了MLM之外，还应用了两个新的辅助正则化器来扩大（a）不同令牌的最后隐藏状态和（b）不同头部的注意力权重之间的差异。这些设计允许模型更清晰地捕捉不同类别的信息，作为减轻表示学习中信息混乱的一种方式，以实现更好的优化。在GLUE基准上的大量实验和烧蚀研究证明了所提出方法的有效性。
Context-Fidelity Boosting: Enhancing Faithful Generation through Watermark-Inspired Decoding
Hunyuan AI Digital Human, Tencent（腾讯文深AI数字人）；McGill University（麦吉尔大学）；Mila - Quebec AI Institute（魁北克人工智能研究所）；Wuhan University（武汉大学）；University of Montreal（蒙特利尔大学）；Tsinghua University（清华大学）；MBZUAI
Position-Aware Depth Decay Decoding ($D^3$): Boosting Large Language Model Inference Efficiency
University of Electronic Science and Technology of China（电子科技大学）；Beijing Academy of Artificial Intelligence（北京人工智能研究院）；Institute of Computing Automation, Chinese Academy of Sciences（中国科学院计算技术研究所）
EquivPruner: Boosting Efficiency and Quality in LLM-Based Search via Action Pruning
School of Computer Science and Technology, University of Science and Technology of China（中国科学技术大学计算机科学与技术学院）；Department of Data Science, City University of Hong Kong（香港城市大学数据科学系）；iFLYTEK Research（iFLYTEK研究院）
Token-Level Policy Optimization: Linking Group-Level Rewards to Token-Level Aggregation via Markov Likelihood
Baidu Inc（百度公司）；College of Computer Science and Technology, Jilin University（吉林大学计算机科学与技术学院）；Key Laboratory of Symbolic Computation and Knowledge Engineering of MOE, Jilin University（吉林大学教育部符号计算与知识工程重点实验室）；State Key Laboratory of AI Safety, Institute of Computing Technology, Chinese Academy of Sciences（中国科学院人工智能安全国家重点实验室）
L-MTP: Leap Multi-Token Prediction Beyond Adjacent Context for Large Language Models
National University of Singapore（新加坡国立大学）；Harbin Institute of Technology（哈尔滨工业大学）；Tsinghua University（清华大学）；Chinese Academy of Sciences（中国科学院）；Central South University（中南大学）
Jakiro: Boosting Speculative Decoding with Decoupled Multi-Head via MoE
Advanced Micro Devices, Inc.（超威半导体公司）；Institute of Artificial Intelligence and Robotics, Xi'an Jiaotong University（西安交通大学人工智能与机器人研究所）；University of HongKong（香港大学）
Beyond Raw Detection Scores: Markov-Informed Calibration for Boosting Machine-Generated Text Detection
Department of Computer Science, Hong Kong Baptist University（香港 Baptist 大学计算机科学系）；School of Software Engineering, South China University of Technology（华南理工大学软件工程学院）；School of Computer Science, University of Science and Technology of China（中国科学技术大学计算机科学学院）
Weighting What Matters: Boosting Sample Efficiency in Medical Report Generation via Token Reweighting
TUM School of Computation, Information and Technology（慕尼黑技术大学计算、信息与技术学院）；Technical University of Munich（慕尼黑技术大学）；Munich Center for Machine Learning (MCML)（慕尼黑机器学习中心）；Department of Computing, Imperial College London（伦敦帝国学院计算机系）
MAIS\&NLPR, Institute of Automation, Chinese Academy of Sciences, Beijing, China（MAIS与NLPR，自动化研究所，中国科学院，北京，中国）；School of Advanced Interdisciplinary Sciences, University of Chinese Academy of Sciences, Beijing, China（先进交叉学科学院，中国科学院大学，北京，中国）
Let the Code LLM Edit Itself When You Edit the Code
National Key Laboratory of General Artificial Intelligence（通用人工智能国家重点实验室）；School of Intelligence Science and Technology, Peking University（北京大学智能科学与技术学院）；ByteDance Inc.（字节跳动公司）
Unlocking Fine-Grained Translation Quality Estimation in LRMs through Mutually Boosting Implicit and Explicit Reasoning
National Key Laboratory for Novel Software Technology, Nanjing University（南京大学新型软件技术国家重点实验室）；Huawei Translation Services Center, Beijing, China（北京华为翻译服务中心）
