The paper " JustiLM: Few-shot Justification Generation for Explainable Fact-Checking of Real-world Claims" addresses the challenge of generating justifications for fact-checking claims. Traditional methods often simplify justification generation by summarizing fact-check articles, which is impractical for new claims without available fact-check articles. The authors propose JustiLM, a novel few-shot justification generation model based on retrieval-augmented language models (RAG). JustiLM leverages fact-check articles as auxiliary resources during training to enhance the quality of generated justifications. They also introduce ExClaim, a new benchmark dataset derived from the WatClaimCheck dataset, containing 6,951 real-world claims and their corresponding veracity labels and human-written justifications. Experiments show that JustiLM outperforms strong baselines, including Flan-T5, Llama2, and the state-of-the-art few-shot RAG model Atlas, and demonstrates promising performance compared to the latest GPT-4 model. Additionally, JustiLM can be extended to jointly predict veracity and generate justifications, improving the veracity prediction task. The paper discusses the limitations and future directions, including the need for human evaluation and the integration of advanced reasoning methods.The paper " JustiLM: Few-shot Justification Generation for Explainable Fact-Checking of Real-world Claims" addresses the challenge of generating justifications for fact-checking claims. Traditional methods often simplify justification generation by summarizing fact-check articles, which is impractical for new claims without available fact-check articles. The authors propose JustiLM, a novel few-shot justification generation model based on retrieval-augmented language models (RAG). JustiLM leverages fact-check articles as auxiliary resources during training to enhance the quality of generated justifications. They also introduce ExClaim, a new benchmark dataset derived from the WatClaimCheck dataset, containing 6,951 real-world claims and their corresponding veracity labels and human-written justifications. Experiments show that JustiLM outperforms strong baselines, including Flan-T5, Llama2, and the state-of-the-art few-shot RAG model Atlas, and demonstrates promising performance compared to the latest GPT-4 model. Additionally, JustiLM can be extended to jointly predict veracity and generate justifications, improving the veracity prediction task. The paper discusses the limitations and future directions, including the need for human evaluation and the integration of advanced reasoning methods.