Zetian Hu

Shunyu Liu, Wenkai Fang, Zetian Hu et al.

Large Language Models (LLMs) have demonstrated unprecedented generative capabilities, yet their alignment with human values remains critical for ensuring helpful and harmless deployments. While Reinforcement Learning from Human Feedback (RLHF) has emerged as a powerful paradigm for aligning LLMs with human preferences, its reliance on complex reward modeling introduces inherent trade-offs in computational efficiency and training stability. In this context, Direct Preference Optimization (DPO) has recently gained prominence as a streamlined alternative that directly optimizes LLMs using human preferences, thereby circumventing the need for explicit reward modeling. Owing to its theoretical elegance and computational efficiency, DPO has rapidly attracted substantial research efforts exploring its various implementations and applications. However, this field currently lacks systematic organization and comparative analysis. In this survey, we conduct a comprehensive overview of DPO and introduce a novel taxonomy, categorizing previous works into four key dimensions: data strategy, learning framework, constraint mechanism, and model property. We further present a rigorous empirical analysis of DPO variants across standardized benchmarks. Additionally, we discuss real-world applications, open challenges, and future directions for DPO. This work delivers both a conceptual framework for understanding DPO and practical guidance for practitioners, aiming to advance robust and generalizable alignment paradigms. All collected resources are available and will be continuously updated at https://github.com/liushunyu/awesome-direct-preference-optimization.

Weijian Xie, Xuefeng Liang, Yuhui Liu et al.

Large Language Models (LLMs) have greatly contributed to the development of adaptive intelligent agents and are positioned as an important way to achieve Artificial General Intelligence (AGI). However, LLMs are prone to produce factually incorrect information and often produce "phantom" content that undermines their reliability, which poses a serious challenge for their deployment in real-world scenarios. Enhancing LLMs by combining external databases and information retrieval mechanisms is an effective path. To address the above challenges, we propose a new approach called WeKnow-RAG, which integrates Web search and Knowledge Graphs into a "Retrieval-Augmented Generation (RAG)" system. First, the accuracy and reliability of LLM responses are improved by combining the structured representation of Knowledge Graphs with the flexibility of dense vector retrieval. WeKnow-RAG then utilizes domain-specific knowledge graphs to satisfy a variety of queries and domains, thereby improving performance on factual information and complex reasoning tasks by employing multi-stage web page retrieval techniques using both sparse and dense retrieval methods. Our approach effectively balances the efficiency and accuracy of information retrieval, thus improving the overall retrieval process. Finally, we also integrate a self-assessment mechanism for the LLM to evaluate the trustworthiness of the answers it generates. Our approach proves its outstanding effectiveness in a wide range of offline experiments and online submissions.

Junjie Zhang, Guozheng Ma, Shunyu Liu et al.

Recent advances in Reinforcement Learning (RL) have underscored its potential for incentivizing reasoning capabilities of Large Language Models (LLMs). However, existing step-level efforts suffer from costly annotations that limit domain coverage, while scalar scores further impose an information bottleneck, offering insufficient semantic bandwidth to improve intermediate decisions. Alternative language-critique approaches, which rely on frozen or external critics, provide richer textual feedback but lack the scalability needed for sustained policy improvement. In this work, we propose language-driven stepwise trajectory redirection, termed as STRIDE, a novel training framework that shifts process supervision from scalar rewards to learnable stepwise language feedback. Specifically, we co-train a generator and a generative verifier using only outcome-based rewards, eliminating external annotations, while delivering sustained policy improvement through jointly aligned verifier training. The verifier's stepwise language critiques explicitly localize and explain failures, enabling the generator to redirect reasoning trajectories at intermediate steps toward alternative decisions. The trajectory redirection design guarantees harmless policy improvement, even under noisy or suboptimal verifier feedback. Experiments on diverse reasoning benchmarks show that STRIDE significantly outperforms state-of-the-art baselines, as well as achieving breakthroughs on zero-pass-rate problems where scalar methods yield no learning signal in our ablation studies, demonstrating the effectiveness of learnable stepwise language feedback for enhancing LLM reasoning.