# Long-Horizon Forecasting of Complete Financial Statements with Forma

arXiv: https://arxiv.org/abs/2608.11327 · HTML: https://mlsanity.com/paper/2608.11327/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 10 of its 43 references that are in the map.
- First posted on arXiv 11 Aug 2026.

## Builds on (most cited first, 10 of 10)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [Are Transformers Effective for Time Series Forecasting?](https://mlsanity.com/paper/2205.13504/) (2022)
- [A decoder-only foundation model for time-series forecasting](https://mlsanity.com/paper/2310.10688/) (2023)
- [Large Language Models Are Zero-Shot Time Series Forecasters](https://mlsanity.com/paper/2310.07820/) (2023)
- [Chronos-2: From Univariate to Universal Forecasting](https://mlsanity.com/paper/2510.15821/) (2025)
- [Are Language Models Actually Useful for Time Series Forecasting?](https://mlsanity.com/paper/2406.16964/) (2024)
- [On the Pitfalls of Heteroscedastic Uncertainty Estimation with Probabilistic Neural Networks](https://mlsanity.com/paper/2203.09168/) (2022)
- [Uncertainty-Aware Lookahead Factor Models for Quantitative Investing](https://mlsanity.com/paper/2007.04082/) (2020)
- [Improving Factor-Based Quantitative Investing by Forecasting Company Fundamentals](https://mlsanity.com/paper/1711.04837/) (2017)
- [Point and probabilistic forecast reconciliation for general linearly constrained multiple time series](https://mlsanity.com/paper/2305.05330/) (2023)

## Abstract (arXiv)

Specialist training beats generalist scale when forecasting financial statements. To our knowledge, no prior work jointly forecasts complete financial statements beyond one year, yet in a discounted-cash-flow valuation most firm value sits past that window. We release ProForma-20Q, a reproducible benchmark for forecasting 78 statement line items 1-20 quarters ahead, for anonymized firms, from past statements and an industry code, scored by change-space R2. On it, Forma, a transformer that reads statements as sets of (account, quarter, value) tuples and maximizes a masked-tuple Gaussian likelihood, beats every competitor we field: classical machine learning, chained gradient boosting, a zero-shot time-series foundation model, and frontier large language models. Its lead widens with horizon, where valuation needs accuracy most, and its Gaussian predictive intervals never under-cover. Forma's forecasts nearly satisfy accounting identities; exact coherence is recoverable at no statistically significant accuracy cost. Its tuple interface supports scenario analysis without retraining, and we show that pinning future revenue paths sharpens the rest of the statement.