New FinATOM model uses language generation for financial prediction

A new research paper proposes FinATOM, a novel approach that uses constrained token generation from a causal language model for financial numerical prediction and allocation. This method aims to unify forecasting and decision-making, moving away from task-specific model heads. In tests using ETF data from 2023-2025, FinATOM demonstrated improvements in pooled Sharpe ratios, outperforming previous methods and showing particular strength in 2025. The research suggests that direct language model token generation is a feasible and promising direction for financial prediction and decision-making across various assets and market conditions. AI

IMPACT This research could lead to more integrated and potentially more effective AI-driven financial forecasting and decision-making systems.

RANK_REASON The cluster contains a research paper detailing a new methodology for financial prediction using language models. [lever_c_demoted from research: ic=1 ai=0.7]

Read on arXiv cs.LG →

• alphaXiv
• arXiv
• CatalyzeX
• DagsHub
• FinATOM
• Gotit.pub
• Hugging Face
• ScienceCast

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →