Financial Numerical Prediction and Allocation as Token Generation

2026-08-10

AI Summary

This work addresses the disconnect between language models and numerical financial targets in traditional forecasting approaches, which rely on task-specific regression or policy heads. The authors propose FinATOM, a headless unified framework that formulates both return prediction and ETF asset allocation as a constrained autoregressive token generation task. Through ordinal pretraining, token-level policy fine-tuning, a DAPO-enhanced GRPO algorithm, and causal mean-variance anchor imitation learning, the model enables end-to-end optimization of the Sharpe ratio. Evaluated on ETF data from 2023 to 2025, FinATOM improves the net Sharpe ratio from 1.394 to 1.494; with multimodal inputs, it achieves an average three-phase Sharpe ratio of 1.540 and, on FinTexTS, delivers cumulative returns of 73.72% and a Sharpe ratio of 2.69.

Abstract

Financial prediction typically relies on task-specific regression, ranking, or policy heads, separating the language model from the numerical object ultimately evaluated. We investigate whether a causal language model can instead represent forecasts and decisions directly through constrained token generation. FinATOM introduces a unified, head-free interface for three-step stock-return forecasting and dynamic five-ETF allocation. The forecasting model autoregressively emits volatility-standardized return tokens and is trained with ordinal and ranking supervision followed by a one-epoch token-level policy stage. The allocation model generates normalized long-only weights; supervised fine-tuning imitates a causal mean--variance anchor, and DAPO-augmented GRPO optimizes realized 21-day Sharpe subject to anchor consistency. In 2023--2025 ETF tests, the allocation policy improves pooled gross Sharpe from 1.428 to 1.529 and net Sharpe under a 5-bp transaction-cost model from 1.394 to 1.494. The multimodal allocation input attains the highest three-period mean Sharpe of 1.540, with its clearest advantage in 2025. On FinTexTS, the SFT and policy strategies achieve 73.52\%/2.68 and 73.72\%/2.69 cumulative-return/Sharpe, respectively. These results support the feasibility of direct language-model token generation for financial numerical prediction and decision-making, while motivating broader tests across assets, regimes, and random seeds.

Problem

• financial prediction
• token generation
• portfolio allocation
• language models
• numerical forecasting

Innovation

• token generation
• financial prediction
• dynamic allocation
• causal language model
• Sharpe ratio optimization