As an emerging education strategy, learnersourcing offers the potential for personalized learning content creation, but also grapples with the challenge of predicting student performance due to inherent noise in student-generated data. While graph-based methods excel in capturing dense learner-question interactions, they falter in cold start scenarios, characterized by limited interactions, as seen when questions lack substantial learner responses. In response, we introduce an innovative strategy that synergizes the potential of integrating Signed Graph Neural Networks (SGNNs) and Large Language Model (LLM) embeddings. Our methodology employs a signed bipartite graph to comprehensively model student answers, complemented by a contrastive learning framework that enhances noise resilience. Furthermore, LLM's contribution lies in generating foundational question embeddings, proving especially advantageous in addressing cold start scenarios characterized by limited graph data interactions. Validation across five real-world datasets sourced from the PeerWise platform underscores our approach's effectiveness. Our method outperforms baselines, showcasing enhanced predictive accuracy and robustness.