中文
相关论文

相关论文: A Mechanistic Understanding of Alignment Algorithm…

200 篇论文

Safety fine-tuning algorithms reduce harmful outputs in language models, yet their mechanisms remain under-explored. Direct Preference Optimization (DPO) is a popular choice of algorithm, but prior explanations, attributing its effects…

机器学习 · 计算机科学 2025-06-10 Yushi Yang , Filip Sondej , Harry Mayne , Andrew Lee , Adam Mahdi

Detoxifying multilingual Large Language Models (LLMs) has become crucial due to their increasing global use. In this work, we explore zero-shot cross-lingual generalization of preference tuning in detoxifying LLMs. Unlike previous studies…

计算与语言 · 计算机科学 2024-11-11 Xiaochen Li , Zheng-Xin Yong , Stephen H. Bach

Recent advances in alignment techniques such as Supervised Fine-Tuning (SFT), Reinforcement Learning from Human Feedback (RLHF), and Direct Preference Optimization (DPO) have improved the safety of large language models (LLMs). However,…

计算与语言 · 计算机科学 2026-02-26 Mengxuan Hu , Vivek V. Datla , Anoop Kumar , Zihan Guan , Sheng Li , Alfy Samuel , Daben Liu

Direct preference optimization (\texttt{DPO}) has emerged as a promising approach for solving the alignment problem in AI. In this paper, we make two counter-intuitive observations about \texttt{DPO}. First, we show that \texttt{DPO} loss…

This study evaluates Direct Preference Optimization (DPO) and its variants for aligning Large Language Models (LLMs) with human preferences, testing three configurations: (1) with Supervised Fine Tuning (SFT), (2) without SFT, and (3)…

计算与语言 · 计算机科学 2025-02-11 Amir Saeidi , Shivanshu Verma , Md Nayem Uddin , Chitta Baral

Direct Preference Optimization (DPO) is an efficient alignment technique that steers LLMs towards preferable outputs by training on preference data, bypassing the need for explicit reward models. Its simplicity enables easy adaptation to…

Recent advancements in Reinforcement Learning with Human Feedback (RLHF) have significantly impacted the alignment of Large Language Models (LLMs). The sensitivity of reinforcement learning algorithms such as Proximal Policy Optimization…

机器学习 · 计算机科学 2025-06-10 Pankayaraj Pathmanathan , Souradip Chakraborty , Xiangyu Liu , Yongyuan Liang , Furong Huang

Existing training-time safety alignment techniques for large language models (LLMs) remain vulnerable to jailbreak attacks. Direct preference optimization (DPO), a widely deployed alignment method, exhibits limitations in both experimental…

计算与语言 · 计算机科学 2025-10-31 Xuandong Zhao , Will Cai , Tianneng Shi , David Huang , Licong Lin , Song Mei , Dawn Song

Recent advancements in open-domain dialogue systems have been propelled by the emergence of high-quality large language models (LLMs) and various effective training methodologies. Nevertheless, the presence of toxicity within these models…

计算与语言 · 计算机科学 2024-05-22 San Kim , Gary Geunbae Lee

Language model (LM) post-training (or alignment) involves maximizing a reward function that is derived from preference annotations. Direct Preference Optimization (DPO) is a popular offline alignment method that trains a policy directly on…

The last year has witnessed the rapid progress of large language models (LLMs) across diverse domains. Among them, CodeLLMs have garnered particular attention because they can not only assist in completing various programming tasks but also…

人工智能 · 计算机科学 2024-10-25 Yibo Miao , Bofei Gao , Shanghaoran Quan , Junyang Lin , Daoguang Zan , Jiaheng Liu , Jian Yang , Tianyu Liu , Zhijie Deng

Recent alignment algorithms such as direct preference optimization (DPO) have been developed to improve the safety of large language models (LLMs) by training these models to match human behaviors exemplified by preference data. However,…

计算与语言 · 计算机科学 2025-03-04 Rheeya Uppaal , Apratim Dey , Yiting He , Yiqiao Zhong , Junjie Hu

Large language models (LLMs) trained on webscale data can produce toxic outputs, raising concerns for safe deployment. Prior defenses, based on applications of DPO, NPO, and similar algorithms, reduce the likelihood of harmful…

The alignment of language models with human preferences is vital for their application in real-world tasks. The problem is formulated as optimizing the model's policy to maximize the expected reward that reflects human preferences with…

计算与语言 · 计算机科学 2024-06-06 Haozhe Ji , Cheng Lu , Yilin Niu , Pei Ke , Hongning Wang , Jun Zhu , Jie Tang , Minlie Huang

Aligning large language models (LLMs) with human preferences has gained significant attention, with Proximal Policy Optimization (PPO) as a standard yet computationally expensive method and Direct Preference Optimization (DPO) as a more…

人工智能 · 计算机科学 2025-02-10 Yuzi Yan , Yibo Miao , Jialian Li , Yipin Zhang , Jian Xie , Zhijie Deng , Dong Yan

Post-alignment of large language models (LLMs) is critical in improving their utility, safety, and alignment with human intentions. Direct preference optimisation (DPO) has become one of the most widely used algorithms for achieving this…

机器学习 · 计算机科学 2025-01-06 Rasul Tutnov , Antoine Grosnit , Haitham Bou-Ammar

In the field of large language models (LLMs), aligning models with the diverse preferences of users is a critical challenge. Direct Preference Optimization (DPO) has played a key role in this area. It works by using pairs of preferences…

计算与语言 · 计算机科学 2024-05-29 Yueqin Yin , Zhendong Wang , Yi Gu , Hai Huang , Weizhu Chen , Mingyuan Zhou

Direct Preference Optimization (DPO) is broadly utilized for aligning Large Language Models (LLMs) with human values because of its flexibility. Despite its effectiveness, it has been observed that the capability of DPO to generate…

机器学习 · 计算机科学 2025-05-20 Wenqiao Zhu , Ji Liu , Lulu Wang , Jun Wu , Yulun Zhang

Direct Preference Optimization (DPO) and its variants have become the de facto standards for aligning large language models (LLMs) with human preferences or specific goals. However, DPO requires high-quality preference data and suffers from…

机器学习 · 计算机科学 2024-11-12 Zhuotong Chen , Fang Liu , Jennifer Zhu , Wanyu Du , Yanjun Qi

The widespread application of large language models (LLMs) raises increasing demands on ensuring safety or imposing constraints, such as reducing harmful content and adhering to predefined rules. While there have been several works studying…

机器学习 · 计算机科学 2026-02-13 Yihan Du , Seo Taek Kong , R. Srikant
‹ 上一页 1 2 3 10 下一页 ›