AI summaryⓘ
The authors explore using a language model to directly generate tokens representing financial forecasts and investment decisions, rather than relying on separate prediction tools. They created FinATOM, which predicts stock returns and allocates investments across ETFs using a unified approach. Their method trains the model to generate return estimates and portfolio weights, improving investment performance measured by Sharpe ratios compared to benchmarks. Tests show this approach works well in both simulated and real ETF trading scenarios. The results suggest that language models can handle numerical financial tasks without needing traditional prediction heads, encouraging further testing in different markets and conditions.
language modelfinancial forecastingETF allocationSharpe ratioautoregressive generationmean-variance optimizationtoken-level policyordinal supervisionranking supervision
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.