Researchers have introduced PRAGMA, a new family of foundation models specifically designed for financial event sequences. This Transformer-based architecture utilizes masked modeling and a self-supervised objective tailored for discrete, variable-length financial records. PRAGMA aims to provide a general-purpose representation layer for banking applications, demonstrating strong performance in downstream tasks like credit scoring, fraud detection, and lifetime value prediction through extracted embeddings and lightweight fine-tuning. AI
IMPACT This model could enable more sophisticated financial analysis and risk management by providing a specialized representation layer for banking data.
RANK_REASON The item is a research paper published on arXiv detailing a new foundation model. [lever_c_demoted from research: ic=1 ai=1.0]
- alphaXiv
- arXiv
- CatalyzeX
- DagsHub
- Gotit.pub
- Hugging Face
- IArxiv
- Revolut
- ScienceCast
- Transformer++
- Vladimir Iashin
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →