Retrieval-Augmented Generation (RAG) effectively mitigates hallucinations and knowledge gaps in Large Language Models (LLMs) for knowledge-intensive tasks by incorporating external web-based knowledge. However, when integrating diverse yet potentially conflicting web-sourced information, RAG systems are prone to knowledge conflicts that manifest as incorrect or inconsistent model behaviors, ultimately leading to unreliable responses. To address this challenge, this paper proposes Conflict-Aware RAG, a general training framework that leverages the model's inherent conflict-sensing capability to build a more robust RAG system via phased optimization. At the core of this framework lies ConScore, a conflict signal that quantifies the model's awareness of potential knowledge conflicts by comparing generative probabilities across distinct knowledge sources. This signal then guides both the construction of training data and a multi-stage optimization workflow: In the Supervised Fine-Tuning (SFT) stage, conflict features are employed to select representative distracting documents, laying the groundwork for core RAG capabilities; in the Direct Preference Optimization (DPO) stage, high-quality preference pairs are constructed using the conflict signal to boost the model's robustness against distracting knowledge; and in the Reranking stage, conflict confidence and information gain are integrated to synergistically optimize the collaboration mechanism between the retriever and LLM. Experiments on six knowledge-intensive question answering (QA) datasets demonstrate that Conflict-Aware RAG significantly outperforms mainstream baselines. Further ablation studies and quantitative analyses validate the method's stability and generalization, laying the foundation for robust RAG systems.

Aspect-based sentiment analysis (ABSA) constitutes a critical subtask within affective computing, whose central challenge involves the accurate and efficient identification of sentiment polarity associated with specific aspect terms in review sentences. Although syntactic knowledge has demonstrated significant benefits in traditional ABSA models, existing approaches based on large language models (LLMs) have largely overlooked such structural information and often fail to comprehensively model both implicit and explicit sentiment expressions. To bridge this gap, we propose TrustSyn, a novel framework designed to enhance LLMs with trustworthy, constituency-structured dependency knowledge for ABSA. Specifically, the input sentences are first parsed using both dependency and constituency parsers. The resulting syntactic information is then restructured into a unified and reliable representation through a trustworthy syntax integration process. This structured knowledge is formalized and injected into LLMs to augment their comprehension of aspect sentiment associations. To the best of our knowledge, this is the first work to integrate constituency-informed dependency structures into LLMs for ABSA. Finally, experimental results demonstrate that TrustSyn consistently outperforms state-of-the art models across five benchmark datasets. Further ablation studies and analyses confirm its robustness and strong generalization capability.