Pith. sign in

REVIEW 3 major objections 6 minor 48 references

Long-Horizon Forecasting of Complete Financial Statements with Forma

T0 review · 3 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Forma, a ~0.9M-parameter tuple-set transformer, jointly forecasts 78 statement line items 1–20 quarters ahead and beats classical ML, chained gradient boosting, a zero-shot foundation model, and frontier LLMs on the new ProForma-20Q…

desk verdict The first credible 78-item, 20-quarter statement-forecasting benchmark with a strong specialist model and an honest protocol; the valuation framing overclaims a dollar-space link it never tests. read the letter →

arxiv 2608.11327 v1 pith:J4CRSSPR submitted 2026-08-11 cs.LG q-fin.CP

Travis L. Johnson , Jiannan Jiang , Soumyabrata Chaudhuri , Yihao Chen , Lauren Falvey , Donal O'Cofaigh This is my paper · ORCID
classification cs.LGq-fin.CP
keywords completefinancialstatementforecastinglong-horizontuple-settransformerchange-spaceR2discountedcashflowvaluationprobabilisticstructuredmissingnessscenarioanalysis
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper argues that forecasting a complete set of financial statements—78 line items, 1 to 20 quarters ahead—is a tractable task for a deliberately small specialist model, and that the task itself deserves a shared benchmark. It introduces ProForma-20Q and Forma, a transformer that reads each firm's history as an unordered set of (account, quarter, value) tuples and learns by predicting masked future values under a Gaussian likelihood. On the benchmark's change-space $R^2$ metric, Forma outperforms every fielded competitor at every horizon beyond two quarters, with the advantage widening as the horizon lengthens toward the five-year window where discounted-cash-flow valuation concentrates most firm value. The paper also shows that the model's predictive intervals are well calibrated, that its forecasts nearly respect accounting identities, and that pinning a known future revenue path improves the forecast of everything else. If these results hold, a sub-million-parameter specialist is more useful than much larger generalist models for the joint, long-horizon statement forecasts that valuation actually needs.

What carries the argument

The load-bearing device is the tuple-set representation combined with masked-tuple Gaussian training. Each reported historical observation becomes a token $(h,id,x)$ with learned account and horizon embeddings plus a learned scalar projection of the standardized value; unreported items contribute no token, and the future is a set of query tokens whose values are hidden. A four-layer transformer encoder with 128-dimensional embeddings attends over this unordered set, and two MLP heads emit a predictive mean and heteroskedastic variance for every queried (item, horizon) pair. Identity-aware masking—when masking touches a complete accounting identity, at least two members are masked—forces the model to learn economics rather than the exact linear algebra of the statements, while a pinned-future masking arm trains it to condition on partial future realizations. Those ingredients, not raw capacity, are what the paper credits for the result: a larger feed-forward network with the same loss underperforms the smaller transformer.

What would settle it

Take the ProForma-20Q test origins, build dollar-space discounted-cash-flow valuations from each model's predictive distributions using the paper's closed form for Gaussian expectations, and rank models by out-of-sample dollar valuation error; if a competitor such as the random forest or an LLM achieves lower dollar-space error than Forma despite losing in change-space $R^2$, the central claim that Forma's lead is what valuation needs would be falsified.

Watch

Extended reading notes

Core claim

ProForma-20Q asks models to predict 78 standardized accounting line items at quarterly horizons 1–20 from 12 quarters of reported history plus an industry code, scoring forecasts by $R^2$ on changes so that persistent levels do not masquerade as skill. Forma encodes the statement as a set of tuples, leaves unreported items absent rather than imputed, masks the future as queries, and trains a heteroskedastic Gaussian head, averaged over five seeds, by maximizing the likelihood of masked values. Its headline result is that on this task a ~0.9M-parameter model beats classical penalized regressions, random forests, feed-forward networks, a re-implemented chained gradient-boosting statement forecaster, a zero-shot time-series foundation model, and three frontier LLMs; the gaps are statistically significant and widen with horizon, reaching at least 3.2 percentage points of change-space $R^2$ by $h=20$. Forma's mixture intervals never under-cover at nominal 50–95% levels, its raw forecasts satisfy accounting identities to a median absolute violation of 3.7% of gross scale, and variance-weighted reconciliation restores exact coherence without a statistically significant squared-error cost. The paper claims this makes complete, distributional, long-horizon statement forecasts—and conditional scenario forecasts from pinned revenue paths—available for the first time as a learned system.

Load-bearing premise

The load-bearing premise is that scoring skill by squared error on standardized quarter-to-quarter changes (change-space $R^2$) is the right objective for valuation; if dollar-accurate, median-targeted, or other losses were used instead, the ranking among models could change.

Editorial extensions

If this is right

  • Long-horizon joint statement forecasts become a practical benchmark task: any model can be scored on the released protocol's common samples, so progress no longer depends only on single-item earnings forecasts.
  • Because the benchmark scores conditional means at 1–20 quarters, a model that wins it supplies the aggregating quantities a DCF valuation needs, and Forma's predictive distribution gives a closed-form route from standardized-space forecasts to expected dollar cash flows.
  • Scenario analysis works without retraining: pinning the true future revenue path lowers pooled MAE from 0.409 to 0.383 and raises change-space $R^2$ from 30.5% to 34.8%, with the gain reaching 7.4 percentage points at $h=20$.
  • Exact accounting coherence can be restored cheaply: a variance-weighted projection drives identity violations to numerical zero with no statistically significant squared-error cost, while equal-weighted reconciliation is catastrophic—so predictive variance is a practical input to reconciliation.
  • Frontier LLMs are not competitive on squared-error skill: even the best LLM underperforms every purpose-trained model except the fade baseline, and its $R^2$ deficit relative to Forma grows from 5.5 percentage points at $h=1$ to 15.9 at $h=20$.
  • The tuple-set formulation is generic: any high-dimensional panel with structured missingness and exact linear constraints—national accounts, energy balances, supply-chain tables—could be attacked with the same missingness-native masking, though the paper only demonstrates accounting statements.
  • Because no firm identity, market, or analyst data is used, the same protocol could be applied to private-firm statements if comparable item-level panels become available.
  • The LLM comparison is conducted with off-the-shelf models at default reasoning settings; fine-tuned or token-matched LLMs might close part of the gap, but the paper does not test that.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 6 minor

Summary. The paper introduces ProForma-20Q, a benchmark for jointly forecasting 78 quarterly financial statement line items at horizons of 1–20 quarters from a 12-quarter reported history plus a Fama-French-48 industry code, with evaluation on changes in origin-frozen, asinh-standardized values. It also proposes Forma, a roughly 0.9M-parameter transformer that represents a statement as a set of (account, quarter, value) tuples, masks future tuples, and is trained with a heteroskedastic Gaussian beta-NLL objective with identity-aware and pinned-future masking. On the benchmark, Forma outperforms classical tabular models, a chained gradient-boosting statement forecaster, Chronos-2, and three frontier LLMs, with its change-space R2 advantage growing from h=3 onward and reaching at least 3.2pp at h=20. The paper further reports that Forma's predictive intervals never under-cover, that exact accounting coherence can be imposed ex post at statistically insignificant R2 cost, and that conditioning on pinned revenue paths improves forecasts of the remaining statement. Code, configurations, and a WRDS-based rebuildable pipeline are released.

Significance. If the central ranking is taken at face value, this is a solid empirical contribution: the tuple-set representation and identity-aware masking are well motivated by the pervasive sparsity of reported statement items, and the evaluation protocol is careful in several respects—temporal splits with target purging, common samples within footprints, quarter-clustered Diebold-Mariano tests, and a clear separation of validation-based model selection from test scoring for the trained models. The release of a reproducible pipeline and trained weights is a genuine asset for the community. The main weakness is that the paper's valuation framing goes beyond what the change-space R2 metric can establish, and the LLM comparison is not as clean as the protocol claims because the better of two prompt arms is selected on the test sample. The benchmark itself and the specialist-versus-generalist result are useful regardless.

major comments (3)
  1. [§3.3, Table 2, Conclusion] The paper's central value proposition—that Forma's long-horizon edge is 'where valuation needs accuracy most'—is not established by the evidence. Change-space R2 on per-item asinh-standardized, clipped values is a legitimate benchmark metric, but no result in the paper connects it to dollar-space DCF valuation error. Because standardization is per-item and origin-specific, equal standardized errors correspond to very different dollar errors across firms and items; and because the predictive distribution is a mixture of Gaussians on clipped values, the closed-form expectation E[sinh(aX+b)] = exp(a^2*sigma^2/2)*sinh(a*mu+b) in §3.3 is not exact at the ±6 clipping boundaries and depends on sigma in a way that R2 does not score. The MAE track (Table 2, Panel B) shows the LLMs within about 4% of Forma on absolute error despite being about 11pp behind on R2, so the ranking is sensitive to loss geometry. I would like to see either a dollar-space evaluation (for example, de-standardized dollar RMSE/MAE on cash-flow items, or the error in a simple DCF valuation built from each model's forecasts) or an explicit reframing of all abstract and conclusion valuation statements as claims about the benchmark R2 metric.
  2. [§5.1, Appendix C.1, §3.2] The protocol states that model-selection decisions use only the validation split, but the LLM comparison selects the better of two prompt arms on the test set: §5.1 says 'Table 2 reports the better arm per model,' and Appendix C.1 lists the unstructured/structured R2 values (for example, 0.186/0.175 for Opus 4.8) and reports the better one. This is test-set model selection for the LLM comparators. It favors the LLMs and therefore does not undermine Forma's win, but it contradicts the paper's 'no test-set hyperparameter selection' claim and weakens the reproducibility of the generalist comparison. Either pre-register a single prompt arm, report both arms with appropriate multiple-comparison handling, or clearly disclose the selection in the protocol description.
  3. [Abstract, §5.1, §6] The headline 'Specialist training beats generalist scale' is broader than the evidence. The generalist comparators are used zero-shot and are not trained or fine-tuned on the forecasting task, so the experiments establish that a small task-trained specialist beats off-the-shelf generalists on ProForma-20Q. They do not establish that specialist training beats generalist scale in general, because a fine-tuned foundation model would combine both. The paper's own limitation paragraph acknowledges this ('only consider off-the-shelf generalists rather than fine-tuned ones'), but the abstract and conclusion still state the broad version. Please either add a fine-tuned or task-trained generalist baseline, or qualify the claim explicitly in the abstract and conclusion.
minor comments (6)
  1. [Table 1] In the Forma row, the '333' under 'Distributional output' appears to be a typographical artifact; it should be a check mark if one is intended.
  2. [§3.3] The definition of change-space R2 is not written as an equation; please state explicitly that R2 = 1 - MSE / Var(realized changes) so readers can see that the denominator is the variance of changes, not levels.
  3. [§5.3, Table 3] The variance-weighted reconciliation cost is described as 'no statistically significant squared-error cost' (R2 drops by 3.8pp, DM t=-1.4), but a 3.8pp drop is economically meaningful; the text should discuss economic as well as statistical significance.
  4. [Appendix C.1] The LLM sample is 2,103 origins from 133 firms out of 321,892 eligible origins; the paper should acknowledge that the LLM comparison is based on a small, possibly non-representative subsample and should report standard errors for the LLM-column metrics.
  5. [§5.2] The parenthetical about LLM pretraining contamination ('comparison is conservatively biased toward the generalists') is important and should appear in the protocol definition rather than only in the results discussion.
  6. [Table 3] The MAE column uses the Gaussian conditional-mean forecasts while Panel B uses the Laplace median-targeting head; the table note explains this, but a reader may misread the two MAE numbers, so a clearer label would help.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the benchmark, model, and evaluation are self-contained; all predictions target held-out test data with origin-frozen standardization.

full rationale

The paper's derivation chain is self-contained. ProForma-20Q defines a temporal split (train 1971–2001, validation 2002–2009, test 2010–2024) with purged targets, so nothing fit or selected before test observes post-2009 outcomes. Standardization parameters (per-item constants, per-quarter means/standard deviations) are estimated from data available at the forecast origin and are origin-frozen, avoiding look-ahead. Forma is trained on train+validation with hyperparameters selected on validation only, then evaluated on the held-out test split. Every competitor is scored on identical common-sample cells within its footprint. The primary metric, change-space R2, is a standard skill measure relative to a no-change baseline; it is not defined in terms of any fitted value. The Gaussian predictive density and the closed-form E[sinh(aX+b)] expression are used for interpretation, not for fitting. No 'prediction' is a renamed fitted parameter, and no load-bearing argument relies on a self-citation or a uniqueness theorem from the authors' prior work. The only self-referential elements are disclosed hyperparameter choices, which is normal empirical ML practice. The skeptic's point about the untested link between change-space R2 and dollar-space DCF valuation is an external-validity concern, not a circularity: the paper does not claim that higher R2 by construction implies better dollar-space valuation; it asserts it as a rationale. Thus the central accuracy claim is not circular.

Assumptions & free parameters 4 free parameters · 3 assumptions · 0 invented entities

The central claim rests on benchmark construction and model hyperparameters. The listed free parameters are fitted or hand-set and are not predicted from first principles. No new physical or economic entities are posited; the tuple-set representation is a modeling artifact, not an invented entity.

free parameters (4)
  • Per-item kurtosis constants k_id (78 values) = 78 values on a 250-point log grid over [1e-2, 1e3], excess kurtosis target 3
    Estimated on training-period observations and frozen (Appendix A.2). They shape the asinh-scaled values and are not predicted from theory.
  • Clip bound = 6 (and 6*sqrt(2) for year-over-year features)
    Section 3.2 and Appendix A.2: standardized inputs and targets are clipped at |x|<=6; chosen by hand.
  • Beta in beta-NLL loss = 0.5
    Section 4 and Appendix B.1: heteroskedastic Gaussian head uses beta-NLL with beta=0.5, from Seitzer et al. [34].
  • Future masking probabilities = 0.5 / 0.95 / 0.1 / 0.05
    Appendix B.1: masking probabilities for full-future, per-tuple, identity-instance, and ungrouped-tuple masking; chosen by hand and not swept.
assumptions (3)
  • domain assumption Accounting identities in Appendix A.5 hold exactly in the Compustat data.
    The benchmark uses these identities for grouped masking and coherence checks; if items are restated or unusual, the identities may not hold and the 'learn economics, not algebra' interpretation weakens.
  • domain assumption Change-space R^2 on standardized values is the appropriate objective for valuation.
    Section 3.3 motivates conditional means and additivity for DCF, but does not validate that this ranking matches dollar-space valuation outcomes; MAE ranking differs materially.
  • domain assumption The 78-item set and 12-quarter lookback are a sufficient conditioning set for 20-quarter forecasts.
    Section 3.1 fixes the information set; omitting macro, analyst, or other firm-level data bounds the achievable forecast skill regardless of model.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Long-Horizon Forecasting of Complete Financial Statements with Forma." pith.science (2026). https://pith.science/paper/J4CRSSPR

@misc{pith2026260811327,
  author       = {Pith},
  title        = {Pith review of: Long-Horizon Forecasting of Complete Financial Statements with Forma},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/J4CRSSPR}},
  note         = {Machine review of arXiv:2608.11327}
}
abstract

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 $R^2$. 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.

Figures

Figures reproduced from arXiv: 2608.11327 by the authors.

Figure 1
Figure 1. Change-space 𝑅 2 by forecast horizon on (a) the full sample (Forma vs. five baselines), (b) the Geert. common sample (Forma vs. the chained GBM), and (c) the LLM common sam￾ple (Forma vs. three frontier LLMs). Markers atop each panel show quarter-clustered DM sig￾nificance at the 5% level against RF, the GBM, and Opus 4.8 (the strongest LLM), respectively († comparator better, ∗ Forma better). 25 [PITH_FULL_IMAGE:f… view at source ↗
Figure 2
Figure 2. Scenario conditioning. Within-model change from pinning the true realized revenue path and forecasting the remaining statement, vs. the unconditional forecast on the same origins: (a) ΔMAE and (b) Δ𝑅2 by horizon, overall and by statement class. 26 [PITH_FULL_IMAGE:figures/full_fig_p027_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

48 extracted references · 35 canonical work pages

  1. [1]

    John Alberg and Zachary C. Lipton. Improving factor-based quantitative investing by fore- casting company fundamentals, 2017. NIPS 2017 Time Series Workshop; v2 revised 26 April 2018

    work page 2017

  2. [2]

    Maddix, Pablo Guerron, Tony Hu, Junming Yin, Nick Erickson, Prateek Mutalik Desai, Hao Wang, Huzefa Rangwala, George Karypis, Yuyang Wang, and Michael Bohlke-Schneider

    Abdul Fatir Ansari, Oleksandr Shchur, Jaris Küken, Andreas Auer, Boran Han, Pedro Mer- cado, Syama Sundar Rangapuram, Huibin Shen, Lorenzo Stella, Xiyuan Zhang, Monon- ito Goswami, Shubham Kapoor, Danielle C. Maddix, Pablo Guerron, Tony Hu, Junming Yin, Nick Erickson, Prateek Mutalik Desai, Hao Wang, Huzefa Rangwala, George Karypis, Yuyang Wang, and Micha...

    arXiv doi:10.48550/arxiv.2510.15821

  3. [3]

    Bernard and Jacob K

    Victor L. Bernard and Jacob K. Thomas. Evidence that stock prices do not fully reflect the implications of current earnings for future earnings.Journal of Accounting and Economics , 13(4):305–340, 1990. doi: 10.1016/0165-4101(90)90008-R

    work page doi:10.1016/0165-4101(90)90008-r 1990

  4. [4]

    Random forests

    Leo Breiman. Random forests. Machine Learning , 45(1):5–32, 2001. doi: 10.1023/A: 1010933404324

    doi:10.1023/a: 2001

  5. [5]

    Missingfinancialdata

    SvetlanaBryzgalova, SvenLerner, MartinLettau, andMarkusPelger. Missingfinancialdata. The Review of Financial Studies, 38(3):803–882, 2025. doi: 10.1093/rfs/hhae036

    work page doi:10.1093/rfs/hhae036 2025

  6. [6]

    Campbell, Harrison Ham, Zhongjin (Gene) Lu, and Katherine Wood

    John L. Campbell, Harrison Ham, Zhongjin (Gene) Lu, and Katherine Wood. Expectations matter: When (not) to use machine learning earnings forecasts.Management Science, 2026. doi: 10.1287/mnsc.2024.05808. ArticlesinAdvance,1–25;publishedonline12February2026

    work page arXiv 2026

  7. [7]

    Uncertainty-awarelookaheadfactor models for quantitative investing

    LakshayChauhan,JohnAlberg,andZacharyC.Lipton. Uncertainty-awarelookaheadfactor models for quantitative investing. InProceedings of the 37th International Conference on Ma- chine Learning, volume 119 ofProceedings of Machine Learning Research , pages 1489–1499. PMLR, 2020. URLhttps://proceedings.mlr.press/v119/chauhan20a.html . 16

    work page 2020

  8. [8]

    Predicting future earnings changes using machine learning and detailed financial data.Journal of Accounting Research, 60(2):467–515, 2022

    Xi Chen, Yang Ha (Tony) Cho, Yiwei Dou, and Baruch Lev. Predicting future earnings changes using machine learning and detailed financial data.Journal of Accounting Research, 60(2):467–515, 2022. doi: 10.1111/1475-679X.12429

    doi:10.1111/1475-679x.12429 2022

Show all 48 references
  1. [9]

    A decoder-only foundation model for time-series forecasting

    Abhimanyu Das, Weihao Kong, Rajat Sen, and Yichen Zhou. A decoder-only foundation model for time-series forecasting. InProceedings of the 41st International Conference on Ma- chine Learning,volume235of Proceedings of Machine Learning Research,pages10148–10167. PMLR, 2024. URLh...

    2024

  2. [10]

    Noise in expectations: Evidence from analyst forecasts

    Tim de Silva and David Thesmar. Noise in expectations: Evidence from analyst forecasts. The Review of Financial Studies, 37(5):1494–1537, 2024. doi: 10.1093/rfs/hhad091

    2024 doi

  3. [11]

    Diebold and Roberto S

    Francis X. Diebold and Roberto S. Mariano. Comparing predictive accuracy. Journal of Business & Economic Statistics, 13(3):253–263, 1995. doi: 10.1080/07350015.1995.10524599

    1995

  4. [12]

    Fore- casting company fundamentals

    Felix Divo, Eric Endress, Kevin Endler, Kristian Kersting, and Devendra Singh Dhami. Fore- casting company fundamentals. Transactions on Machine Learning Research , 2025. ISSN 2835-8856. URL https://openreview.net/forum?id=haf78jerSt

    2025

  5. [13]

    Fama and Kenneth R

    Eugene F. Fama and Kenneth R. French. Industry costs of equity.Journal of Financial Eco- nomics, 43(2):153–193, 1997. doi: 10.1016/S0304-405X(96)00896-3

    1997 doi

  6. [14]

    Quarterly accounting data: Time-series properties and predictive-ability results

    George Foster. Quarterly accounting data: Time-series properties and predictive-ability results. The Accounting Review, 52(1):1–21, 1977. doi: 10.2308/tar-4493420

    1977 doi

  7. [15]

    Projecting financial statements with chained machine learning

    Paul Geertsema, Helen Lu, and Guang Ma. Projecting financial statements with chained machine learning. SSRN Working Paper 5039433, 2026. doi: 10.2139/ssrn.5039433

    2026 doi

  8. [16]

    Point and probabilistic forecast reconciliation for general linearly constrained multiple time series.Statistical Methods & Applications, 33 (2):581–607, 2024

    Daniele Girolimetto and Tommaso Di Fonzo. Point and probabilistic forecast reconciliation for general linearly constrained multiple time series.Statistical Methods & Applications, 33 (2):581–607, 2024. doi: 10.1007/s10260-023-00738-6. 17

    2024 doi

  9. [17]

    Tilmann Gneiting and Adrian E. Raftery. Strictly proper scoring rules, prediction, and estimation. Journal of the American Statistical Association , 102(477):359–378, 2007. doi: 10.1198/016214506000001437

    2007 doi

  10. [18]

    Thecontin- uousrankedprobabilityscoreforcircularvariablesanditsapplicationtomesoscaleforecast ensembleverification

    EricP.Grimit,TilmannGneiting,VeronicaJ.Berrocal,andNicholasA.Johnson. Thecontin- uousrankedprobabilityscoreforcircularvariablesanditsapplicationtomesoscaleforecast ensembleverification. Quarterly Journal of the Royal Meteorological Society,132(621C):2925– 2942, 2006. doi: 10.1...

    2006 doi

  11. [19]

    Léo Grinsztajn, Edouard Oyallon, and Gaël Varoquaux. Why do tree-based models still outperform deep learning on typical tabular data? InAdvances in Neural Information Pro- cessing Systems 35 (NeurIPS 2022), Datasets and Benchmarks Track , pages 507–520. Curran Associates, Inc....

    2022 doi

  12. [20]

    Large language models are zero-shot time series forecasters

    Nate Gruver, Marc Finzi, Shikai Qiu, and Andrew Gordon Wilson. Large language models are zero-shot time series forecasters. InAdvances in Neural Information Processing Systems , volume 36, pages 19622–19635, 2023. doi: 10.52202/075280-0861

    2023 doi

  13. [21]

    Testing the equality of prediction mean squared errors.International Journal of Forecasting, 13(2):281–291, 1997

    David Harvey, Stephen Leybourne, and Paul Newbold. Testing the equality of prediction mean squared errors.International Journal of Forecasting, 13(2):281–291, 1997. doi: 10.1016/ S0169-2070(96)00719-4

    1997

  14. [22]

    Kelly, and Lasse Heje Pedersen

    Oliver Hellum, Theis Ingerslev Jensen, Bryan T. Kelly, and Lasse Heje Pedersen. The power of the common task framework. SSRN Working Paper 5242901 , 2025. doi: 10.2139/ssrn. 5242901

    2025 doi

  15. [23]

    Interpretable machine learning for earnings forecasts: Leveraging high-dimensional financial statement data.SSRN Working Paper 4619313, 2026

    Dieter Hess, Frederik Simon, and Sebastian Weibels. Interpretable machine learning for earnings forecasts: Leveraging high-dimensional financial statement data.SSRN Working Paper 4619313, 2026. doi: 10.2139/ssrn.4619313. CFR Working Paper 25-06

    2026 doi

  16. [24]

    van Dijk, and Yinglei Zhang

    Kewei Hou, Mathijs A. van Dijk, and Yinglei Zhang. The implied cost of capital: A new 18 approach. Journal of Accounting and Economics, 53(3):504–526, 2012. doi: 10.1016/j.jacceco. 2011.12.001

    2012 doi

  17. [25]

    Machinelearningandthepredic- tion of changes in profitability.Contemporary Accounting Research, 40(4):2643–2672, 2023

    StewartJones,WilliamJ.Moser,andMatthewM.Wieland. Machinelearningandthepredic- tion of changes in profitability.Contemporary Accounting Research, 40(4):2643–2672, 2023. doi: 10.1111/1911-3846.12888

    2023

  18. [26]

    Theodosia Konstantinidi and Peter F. Pope. Forecasting risk in earnings. Contemporary Accounting Research, 33(2):487–525, 2016. doi: 10.1111/1911-3846.12158

    2016

  19. [27]

    Kosiorek, Seungjin Choi, and Yee Whye Teh

    Juho Lee, Yoonho Lee, Jungtaek Kim, Adam R. Kosiorek, Seungjin Choi, and Yee Whye Teh. Set Transformer: A framework for attention-based permutation-invariant neural networks. In Proceedings of the 36th International Conference on Machine Learning , vol- ume 97 ofProceedings of...

    2019

  20. [28]

    Li and Partha Mohanram

    Kevin K. Li and Partha Mohanram. Evaluating cross-sectional forecasting models for im- plied cost of capital. Review of Accounting Studies , 19(3):1152–1185, 2014. doi: 10.1007/ s11142-014-9282-y

    2014

  21. [29]

    Newey and Kenneth D

    Whitney K. Newey and Kenneth D. West. A simple, positive semi-definite, heteroskedas- ticity and autocorrelation consistent covariance matrix.Econometrica, 55(3):703–708, 1987. doi: 10.2307/1913610

    1987 doi

  22. [30]

    Doron Nissim and Stephen H. Penman. Ratio analysis and equity valuation: From research to practice.Review of Accounting Studies, 6(1):109–154, 2001. doi: 10.1023/A:1011338221623

    2001 doi

  23. [31]

    Probabilisticforecastreconciliation: Properties,evaluationandscoreoptimisation

    AnastasiosPanagiotelis,PuwasalaGamakumara,GeorgeAthanasopoulos,andRobJ.Hynd- man. Probabilisticforecastreconciliation: Properties,evaluationandscoreoptimisation. Eu- ropean Journal of Operational Research, 306(2):693–706, 2023. doi: 10.1016/j.ejor.2022.07.040. 19

    2023 doi

  24. [32]

    Analysts’ forecasting models and uncertainty about the past

    Min Park and Tzachi Zach. Analysts’ forecasting models and uncertainty about the past. Review of Accounting Studies, 30(3):2376–2418, 2025. doi: 10.1007/s11142-025-09898-0

    2025 doi

  25. [33]

    Werner, Konstantinos Benidis, Pedro Mercado, Jan Gasthaus, and Tim Januschowski

    Syama Sundar Rangapuram, Lucien D. Werner, Konstantinos Benidis, Pedro Mercado, Jan Gasthaus, and Tim Januschowski. End-to-end learning of coherent probabilistic forecasts for hierarchical time series. InProceedings of the 38th International Conference on Machine Learning, vol...

  26. [34]

    On the pitfalls of heteroscedastic uncertainty estimation with probabilistic neural networks

    Maximilian Seitzer, Arash Tavakoli, Dimitrije Antić, and Georg Martius. On the pitfalls of heteroscedastic uncertainty estimation with probabilistic neural networks. InInternational Conference on Learning Representations, 2022. URL https://openreview.net/forum? id=aPOpXlnV1T

    2022

  27. [35]

    Eric C. So. A new approach to predicting analyst forecast errors: Do investors overweight analyst forecasts? Journal of Financial Economics , 108(3):615–640, 2013. doi: 10.1016/j. jfineco.2013.02.002

    2013 doi

  28. [36]

    Merrill, Vinayak Gupta, Tim Althoff, and Thomas Hartvigsen

    Mingtian Tan, Mike A. Merrill, Vinayak Gupta, Tim Althoff, and Thomas Hartvigsen. Are language models actually useful for time series forecasting? InAdvances in Neural Informa- tion Processing Systems, volume 37, pages 60162–60191, 2024. doi: 10.52202/079017-1922

    2024 doi

  29. [37]

    Attention is all you need

    AshishVaswani, NoamShazeer, NikiParmar, JakobUszkoreit, LlionJones, AidanN.Gomez, Łukasz Kaiser, and Illia Polosukhin. Attention is all you need. InAdvances in Neural In- formation Processing Systems, 2017. URLhttps://proceedings.neurips.cc/paper_ files/paper/2017/hash/3f5ee24...

    2017

  30. [38]

    Waggoner and Tao Zha

    Daniel F. Waggoner and Tao Zha. Conditional forecasts in dynamic multivariate models. The Review of Economics and Statistics , 81(4):639–651, 1999. doi: 10.1162/003465399558508. 20

    1999 doi

  31. [39]

    Wahlen, Stephen P

    James M. Wahlen, Stephen P. Baginski, and Mark T. Bradshaw.Financial Reporting, Fi- nancial Statement Analysis and Valuation . Cengage Learning, 10th edition, 2023. ISBN 9780357722091

    2023

  32. [40]

    Wickramasuriya, George Athanasopoulos, and Rob J

    Shanika L. Wickramasuriya, George Athanasopoulos, and Rob J. Hyndman. Optimal forecast reconciliation for hierarchical and grouped time series through trace minimiza- tion. Journal of the American Statistical Association , 114(526):804–819, 2019. doi: 10.1080/ 01621459.2018.1448825

    2019

  33. [41]

    Aretransformerseffectivefortimeseries forecasting? InProceedings of the AAAI Conference on Artificial Intelligence,volume37,pages 11121–11128, 2023

    AilingZeng,MuxiChen,LeiZhang,andQiangXu. Aretransformerseffectivefortimeseries forecasting? InProceedings of the AAAI Conference on Artificial Intelligence,volume37,pages 11121–11128, 2023. doi: 10.1609/aaai.v37i9.26317. 21 Table 1: Academic research on forecasting firm-level ...

    2023 doi

  34. [43]

    [12]) Five selected items across all three statements 1 or 5 1 24 models 𝑑 Varies Varies — Chained statements (Geertsema et al

    — Multi-model fundamentals benchmark (Divo et al. [12]) Five selected items across all three statements 1 or 5 1 24 models 𝑑 Varies Varies — Chained statements (Geertsema et al. [15]) Partial income statement and balance sheet 29 + 19𝑏 1 Chained GBMs Sequential chain — Articul...

    1970

  35. [44]

    Revenue (revtq) first. Anchor on the SAME fiscal quarter a year earlier: revtq[t] ~= revtq[t-4] x (1 + g), where g is a year-over-year growth rate read from the recent trailing-4-quarter trend and FADED toward a modest long-run rate across the 20-quarter horizon -- do not let ...

  36. [45]

    Operating lines as ratios to revenue. Project cogsq, xsgaq (which includes xrdq), dpq, and stkcoq each as a share of revtq, held near its trailing-4-quarter median and drifting only with a clear operating- leverage reason; convert back to a level = ratio x revtq[t]. Income tax...

  37. [46]

    Working capital via turnover. Project receivables (rectq) off revtq (days-sales-outstanding) and inventory (invtq) and payables (apq) off cogsq (days-inventory / days-payable); hold each turnover near its trailing level, then convert back to a level using projected revtq/cogsq

  38. [47]

    ppentq[t] ~= ppentq[t-1] + capxq - dpq, with capxq projected as a share of revtq

    PP&E by roll-forward. ppentq[t] ~= ppentq[t-1] + capxq - dpq, with capxq projected as a share of revtq

  39. [48]

    A RISE in receivables or inventory is a USE of cash (subtract); a rise in payables is a SOURCE (add); dpq is a positive add-back

    Cash flow first, then SET cash and retained earnings by roll-forward -- do NOT forecast cheq or req independently: * Operating cash flow must be consistent with earnings and working capital: oancfq ~= niq + dpq - Delta rectq - Delta invtq + Delta apq. A RISE in receivables or ...

  40. [2021]

    URL https://proceedings.mlr.press/v139/rangapuram21a.html

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.