Continuous Uniqueness and Novelty Metrics for Generative Modeling of Inorganic Crystals

2025-10-14

AI Summary

Existing uniqueness and novelty metrics for inorganic crystal generation rely on distance functions suffering from four critical flaws: (1) inability to quantify structural similarity, (2) conflation of compositional and structural differences, (3) lack of Lipschitz continuity, and (4) failure to be invariant under atomic permutations. To address these, we propose a novel, theoretically grounded framework for uniqueness and novelty assessment based on Lipschitz-continuous distance functions. Our approach is the first to decouple compositional and structural discrepancies while rigorously guaranteeing permutation invariance. We design and integrate two continuous crystal distance functions—enabling fine-grained similarity comparison and statistical analysis. Experiments demonstrate that our metric uncovers chemically meaningful structural patterns in material space overlooked by conventional metrics (e.g., FCD, COV), significantly improving assessment stability, discriminative power, and physical interpretability. This provides a more reliable, mathematically consistent benchmark for evaluating generative materials models.

Technology Category

• Search and Optimization: Mixed Discrete/Continuous Search
• Constraint Satisfaction and Optimization: Other Foundations of Constraint Satisfaction
• Machine Learning: Calibration & Uncertainty Quantification

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Security and Privacy: Large-scale security measurements

Abstract

To address pressing scientific challenges such as climate change, increasingly sophisticated generative artificial intelligence models are being developed that can efficiently sample the large chemical space of possible functional materials. These models can quickly sample new chemical compositions paired with crystal structures. They are typically evaluated using uniqueness and novelty metrics, which depend on a chosen crystal distance function. However, the most prevalent distance function has four limitations: it fails to quantify the degree of similarity between compounds, cannot distinguish compositional difference and structural difference, lacks Lipschitz continuity against shifts in atomic coordinates, and results in a uniqueness metric that is not invariant against the permutation of generated samples. In this work, we propose using two continuous distance functions to evaluate uniqueness and novelty, which theoretically overcome these limitations. Our experiments show that these distances reveal insights missed by traditional distance functions, providing a more reliable basis for evaluating and comparing generative models for inorganic crystals.

Problem

• Overcoming limitations in crystal distance functions for generative models
• Quantifying similarity between inorganic crystal compounds more reliably
• Providing continuous uniqueness and novelty metrics for materials evaluation

Innovation

• Continuous distance functions overcome traditional metric limitations
• New metrics quantify similarity and distinguish compositional differences
• Proposed approach ensures Lipschitz continuity and permutation invariance