Reproduce “ValueGraph: Value-Signal Guided Graph Pre-training for Contextualized User Representation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yitong Han, Wei Gao, Yi Zhao, Prasanta Bhattacharya, Fengzhu Zeng, Mohammad Amanlou
ValueGraph learns user representations from post text and reply graphs, using masked graph pre-training and noisy MoralBERT signals averaged across posts to guide contrastive and clustering objectives. On conversational stance and bot-detection benchmarks, it improves on tested baselines; on RumourEval19, it reaches 0.77 accuracy and 0.71 macro-F1 versus ModernBERT’s 0.74 and 0.59. These gains do not validate inferred values as psychological profiles.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.