Results show that post-training can substantially improve financial forecasting performance, together with changes in how the model investigates the market, on this outcome-selected benchmark.
Abstract
Post-training has been shown to significantly improve language models'performance on tasks with verifiable outcomes, including mathematical reasoning, software engineering, and computer use. However, whether the same approach can improve forecasting in financial markets is much less clear. Compared with tasks with verifiable outcomes, not only are realized returns noisy, but even what constitutes a relevant information set for making effective predictions is not obvious a priori: the model must decide which observations to gather and then commit to a numerical judgment before the outcome is known. We study this question in a chronological stock-price sandbox, where a language model gathers price, volume, relative-performance, and market-context evidence and predicts a future return. We post-train Qwen3-4B with supervised fine-tuning (SFT) on tool-use demonstrations, then proximal policy optimization (PPO) with a terminal reward given by the forecast score against the realized return. The resulting AURA-4B more than doubles the starting direction--magnitude score, from 20.94 to 43.31, and is comparable to frontier language models on this benchmark. Conditional magnitude agreement rises from 33.3 to 66.2, while directional accuracy changes from 62.9 to 65.4. SFT expands tool use, and PPO further increases the share of ranking and market-context queries. These results show that post-training can substantially improve financial forecasting performance, together with changes in how the model investigates the market, on this outcome-selected benchmark.
An agentic forecasting environment, dataset, and harness built from 2,100+ resolved Polymarket questions is introduced; the agent acquires its own context at rollout time (web search, page reading, and financial time series, all restricted by layered leak filtering to information published before each question's cutoff...
Yusuf Afifi, Artur Kiulian, Anton Polishko et al.· 0 citations
The 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.
Investment institutions exert substantial influence on asset prices, liquidity, and market stability, making the ability to forecast their portfolio adjustments both academically and practically important. We study modeling investment institutions by fine-tuning large language models (LLMs) to predict next-quarter chan...
Yu-Xiang Cheng, Nan-Jiang Du, Qi-Da Mou et al.· IEEE Conference on Computati...· 0 citations
A competence gate is introduced that estimates domain-level source weights from resolved outcomes, shrinks uncertain estimates toward a global weight, and recalibrates the pooled forecast and provides a practical approach for selective model use based on measured marginal value.
Forecasting short-term movements in financial markets remains challenging because market prices are influenced by rapidly changing economic conditions, investor sentiment, and volatility. While machine-learning methods have demonstrated promise for financial forecasting, many forecasting tools remain difficult for non-...
C. Ibebuchi· Proceedings of the 7th Natio...· 0 citations
Threshold and judge diagnostics further clarify the limits of interpreting realization as precise anticipation and provides a common task for studying which research ideas a community subsequently pursues and how reliably this outcome can be measured.
Fenghai Li, Zi-Han Tang, Hao-Fei Yu et al.· 1 citation