A minimal implementation of Drifting Models for 2D toy data. Unlike diffusion/flow models that iterate at inference, drifting models evolve the pushforward distribution during training and generate in a single forward pass (1-NFE). The drifting field V governs sample movement: V -> 0 as generated matches data.

Minimal PyTorch implementation of “Generative Modeling via Drifting” with visual diagnostics on 2D benchmarks.

🌀 Implement minimal drifting models for generative modeling with one-step training and efficient inference without score functions or iteration.

Minimal reproducible sandbox for the Linghe-Core Personality Model drift test

Development of 'Prêt à dépenser', a credit scoring and predictive analytics tool for Home Credit, aimed at assessing loan repayment likelihood for clients with minimal credit history, featuring a logistic regression model, data drift monitoring, and an interactive Streamlit dashboard for visualization.