Zhendong Mao

In the era of Large Language Models (LLMs), the Internet is flooded with LLM-generated content, including both aligned LLM-generated content that contains equivalent facts to human-written content and distorted LLM-generated content with factual discrepancies. However, there is a significant gap in understanding and addressing the impact of such content on Retrieval-Augmented Generation (RAG) systems, which rely on accurate retrieved knowledge to generate response. This article comprehensively evaluate the impact of aligned and distorted LLM-generated content on the retrieval and generation stages of RAG systems. Notably, we reveal that distorted LLM-generated content is not only more likely to be retrieved but also significantly degrades generation quality, whereas aligned content can, in some cases, enhance it. To mitigate this, we propose a factual consistency-aware adaptive filtering approach to selectively filter out distorted LLM-generated content from retrieved documents. Experimental results demonstrate that our method significantly improves generative performance in scenarios that mix with LLM-generated content and is broadly applicable across various RAG systems.

Reward modeling is essential for aligning Large Language Models(LLMs) with human preferences, yet conventional reward models suffer from poor interpretability and heavy reliance on costly expert annotations. While recent rubric-based approaches enhance evaluation transparency, they lack systematic quality control, yielding noisy and redundant criteria, failing to mitigate persistent biases (e.g., verbosity, position) in LLM evaluators, and creating a scalability-reliability trade-off. To address these limitations, we propose CDRRM (Contrast-Driven Rubric Reward Model), a framework built on a novel Contrast-then-Synthesis paradigm for high-quality rubric generation and guided preference judgment. CDRRM first conducts multi-dimensional contrastive profiling on preference pairs to identify causal discriminative factors, then synthesizes these insights into compact, context-aware rubrics to guide preference judgments. Extensive experiments on three authoritative benchmarks (RewardBench, RMBench, RMB) demonstrate that CDRRM achieves state-of-the-art performance across diverse domains and effectively mitigates aforementioned evaluation biases. Notably, our approach delivers exceptional data efficiency: training the rubric generator on only 3k high-quality samples empowers a frozen pre-trained judge model to outperform fully fine-tuned baselines. This work offers a scalable, interpretable, and data-efficient path for reward modeling.

Retrieval-Augmented Generation (RAG) is an effective method to enhance the capabilities of large language models (LLMs). Existing methods typically optimize the retriever or the generator in a RAG system by directly using the top-$k$ retrieved documents. However, two key issues inherent in the training data constrain the effectiveness of this training paradigm: (1) across different queries, the top-$k$ retrieved documents vary greatly in content quality, with some providing valuable knowledge while others lack critical information or are even misleading, and training on such data in a purely random manner may impair the generator's ability to extract key information; (2) for a given query, the limited set of $k$ documents often exhibits low discriminability, and training solely on them makes it difficult for the retriever to learn how to distinguish between relevant and irrelevant documents. To address these issues, we introduce DACL-RAG, a multi-stage RAG training framework that combines a multi-level Data Augmentation strategy with a multi-stage Curriculum Learning paradigm. The data augmentation strategy constructs comprehensive and diverse training sets with controllable difficulty levels through sample evolution, while the curriculum learning paradigm organizes them into progressive stages for training, ensuring stable and consistent improvements, thereby optimizing the overall performance and generalization of the RAG system more effectively. Our DACL-RAG framework demonstrates consistent effectiveness across four open-domain QA datasets, achieving performance gains of 2% to 4% over multiple advanced methods.