Researchers have developed a system called AViD Journal that can verify whether an automatically generated theorem in mathematics is new or already known. The system uses a decision tree to evaluate three dimensions: whether the theorem has been previously formalized, its non-triviality, and how similar it is to existing proofs. In testing, AViD Journal identified several limitations of this approach, including the fact that successful compilation does not guarantee semantic fidelity, and the recall ceiling is imposed by the coverage of theorem indices rather than the similarity metric. --- Why it matters: This matters because AI systems can currently verify the correctness of math theorems but not their novelty. AViD Journal's ability to identify new results could improve the efficiency of mathematical research and help prevent duplicate publications. Source: https://arxiv.org/abs/2608.14669