Pith. sign in

REVIEW 3 major objections 5 minor 72 references

The Evolution of Alpha in Finance Harnessing Human Insight and LLM Agents

T0 review · 3 major / 5 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 Alpha investing now maps to five stages, from intuition to LLM agents.

desk verdict A useful survey and maturity taxonomy weighed down by Section 6.1's unvalidated, internally inconsistent metrics; on balance it deserves peer review rather than desk rejection. read the letter →

arxiv 2505.14727 v1 pith:WFNRSIRO submitted 2025-05-20 cs.LG q-fin.CP

Mohammad Rubyet Islam This is my paper · ORCID
classification cs.LGq-fin.CP
keywords alphagenerationLLMagentsfive-stagetaxonomymultimodallearningfinancialmachinetrustscoreexplainabilityAIgovernanceinfinance
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 review paper argues that the evolution of alpha generation—returns above market benchmarks—can be organized into a single five-stage taxonomy: manual/fundamental strategies, statistical models, classical machine learning, deep learning, and LLM-driven agents. The claim is that this taxonomy gives practitioners a unified way to assess a system's maturity by its level of automation, modeling complexity, and decision intelligence. The paper also proposes a composite Trust Score and a SHAP-weighted explainability metric to evaluate interpretability and reliability alongside predictive performance. A sympathetic reader would care because, if correct, the framework turns a fragmented field into a common language for benchmarking, infrastructure alignment, and AI governance in finance.

What carries the argument

The central object is the five-stage taxonomy of alpha strategy evolution, a classification that maps manual heuristics, statistical models, classical ML, deep learning, and LLM-based agents onto a single maturity ladder. The taxonomy does the work of turning scattered modeling trends into a benchmarkable progression, emphasizing automation and decision intelligence as the organizing axes. Alongside it, the paper introduces two evaluation formulas: a composite Trust Score (Eq. 3), which sums weighted components for attribution consistency, output stability, factual correctness, and domain alignment, and a SHAP-weighted explainability metric (Eq. 4), which aggregates absolute SHAP values weighted by feature importance to indicate whether high-influence features are also transparent. These metrics are meant to operationalize governance and compliance readiness for AI-driven alpha systems.

What would settle it

A concrete test would be to survey a sample of live institutional alpha systems and classify each by the taxonomy's five stages, then check whether any well-performing system uses a mix of stages with no single dominant method; if such hybrid systems are common and their performance is not explained by the nearest stage, the taxonomy's ordering claim would fail. Alternatively, computing the proposed Trust Score for two models that are deployed in practice and known to have opposite compliance track records—and finding that the score does not rank them correctly—would falsify the metric's usefulness.

Watch

Extended reading notes

Core claim

The paper's central claim is that the history and future of alpha generation strategies can be captured by a five-stage maturity taxonomy, ordered as (1) manual and fundamental alpha, (2) statistical alpha, (3) classical machine learning alpha, (4) deep learning alpha, and (5) agentic alpha powered by large language models. Each stage is defined by increasing automation, modeling complexity, and decision-making intelligence, with the later stages emphasizing end-to-end representation learning, multimodal data fusion, and autonomous tool-using agents. The paper further introduces two quantitative evaluation tools: a composite Trust Score (Eq. 3) that aggregates attribution, stability, factuality, and alignment, and a SHAP-weighted explainability metric (Eq. 4) that measures whether the most influential features are the most interpretable. The overall claim is that these instruments together provide a unified framework for assessing maturity, aligning infrastructure, and guiding the responsible deployment of next-generation alpha systems.

Load-bearing premise

The taxonomy assumes that alpha strategies evolve through five separable, linearly ordered stages and that a system's maturity can be read from its dominant modeling method, yet real production books often combine human discretion, factor models, ML, and LLM tooling at the same time, which would break a strict stage ordering.

Editorial extensions

If this is right

  • Practitioners could use the taxonomy as a maturity checklist: locating a system's dominant modeling stage would reveal which infrastructure, governance, and talent investments matter most.
  • The Trust Score and explainability metric could serve as a pre-deployment governance screen, complementing backtest performance with a standardized interpretability and reliability score.
  • The taxonomy gives regulators and auditors a shared vocabulary for discussing automated trading systems, potentially easing compliance conversations around model logic and oversight.
  • The paper's stage ordering implies that agentic LLM systems are currently best positioned as decision-support co-pilots rather than fully autonomous trading agents, given hallucination, latency, and governance challenges.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The strict linear ordering is likely a simplification: real production books often combine human discretion, factor models, ML signals, and LLM tooling simultaneously, so the taxonomy may be more useful as a capability map than as a strict maturity ladder.
  • The Trust Score's weights are left tunable, which makes it flexible but also means its validity will depend on whether institutions can agree on calibration; an implicit testable extension is whether higher Trust Scores actually predict better audit outcomes or fewer compliance incidents.
  • A natural next step the paper does not develop is empirically validating whether the five stages correlate with measurable performance or risk attributes, such as lower drawdowns or higher information ratios, across real funds.
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 / 5 minor

Summary. This paper is a literature review and taxonomy proposal that traces the evolution of alpha generation from manual, intuition-driven strategies through statistical models, classical machine learning, deep learning, and agentic systems powered by large language models (LLMs). It introduces a five-stage maturity taxonomy in Section 5, discusses cross-cutting challenges such as interpretability, data fragility, governance, and regulatory compliance, and proposes two quantitative evaluation tools: a composite Trust Score in Equation (3) and a SHAP-weighted explainability metric in Equation (4). The paper contains no empirical tests; its contributions are presented as a descriptive framework and a set of proposed metrics for model governance.

Significance. If the framework were validated, it would provide practitioners with a common language for discussing maturity of alpha-generation systems and a starting checklist for model governance. The paper does a useful service by synthesizing a broad literature across quantitative finance, deep learning, and LLM agents, and by explicitly connecting technical evolution to regulatory and compliance concerns. The collection of references is extensive, and the system-level perspective is a reasonable complement to model-centric surveys. However, the quantitative contributions are not derived or validated, and Equation (4) does not actually measure explainability in the sense claimed. The paper is therefore best regarded as an organizing survey plus a set of proposals, not as a tested assessment framework. Its practical value would be strengthened substantially by worked examples, defined scales, and a comparison against existing interpretability metrics.

major comments (3)
  1. [§6.1, Eq. (3)] The metric E is defined as E = Σ|ω_i · SHAP_i| and is claimed to indicate that 'features with greater predictive influence are also more explainable.' This claim does not follow from the equation under either reading of ω_i. If ω_i is, as stated in the same paragraph, 'the relative importance weight assigned to feature i,' then E is simply a weighted L1 norm of SHAP values: it increases whenever a high-SHAP feature is present, regardless of whether that feature is interpretable, auditable, or economically meaningful. If ω_i is instead intended as an interpretability weight, the paper provides no definition, scale, or construction for it, and the sentence becomes self-contradictory. Since this metric is presented as the paper's quantitative contribution to explainability and compliance readiness, this internal inconsistency is load-bearing for the central governance claim.
  2. [§6.1, Eq. (3)] The Trust Score in Equation (3) is defined as a weighted sum of Attribution_i, Stability_i, Factuality_i, and Alignment_i, but none of these components is given a measurable definition, a normalization scale, or an aggregation rule. The weights ω_1 through ω_4 are described as 'tunable' without any guidance on how to set them or interpret the resulting score, which makes the metric non-operational. Additionally, the third term is textually garbled: it reads '(Stability * Factuality_i)', mixing an unindexed 'Stability' with the indexed 'Factuality_i', so the expression is ambiguous even as notation. Because the paper claims this is a 'standardized, modular framework' for assessing model readiness and deployability, the absence of definitions prevents verification or practical use.
  3. [§5] The five-stage taxonomy is asserted as a linear, separable ordering of alpha-generation methods, but no criteria are given for assigning a real-world system to a single stage, and hybrid architectures—where discretionary human oversight, factor models, ML predictors, and LLM tooling coexist—are not addressed. The paper presents the taxonomy as a basis for 'evaluating maturity, aligning infrastructure, and guiding responsible development,' yet no empirical evidence, case studies, or worked applications are provided to show that the stage ordering is distinctive or useful for assessment rather than merely descriptive. This is a methodological gap for a central load-bearing claim of the paper.
minor comments (5)
  1. [§6.1] There are several typographical errors in the introduction of Equation (4), including 'derived fro, the cocept of exaplainalble AI'; these should be corrected.
  2. [Table 4] The title of Table 4 reads 'Deep Learning Modles for Alpha Signal Pipelines'; 'Modles' should be 'Models'.
  3. [§7.1] Section 7.1 states that 'Foundational methods are detailed in Section 4.2,' but the multimodal fusion methods referenced here are actually discussed in Section 3.2; the cross-reference should be corrected.
  4. [References] The reference list has inconsistent formatting; for example, reference [42] uses an unusual 'Zhang, Y., Zhang, J., & Qin, Z. (2021)' style, and several arXiv citations lack version numbers. A careful editorial pass is needed.
  5. [Author Information] The author biography at the end of the manuscript is lengthy and self-promotional; for a technical paper, it should be reduced to a single affiliation and contact line or removed.

Circularity Check

2 steps flagged · score 4.0 of 10

The SHAP-weighted explainability metric in Eq. 4 is a renamed attribution-magnitude score, and Eq. 5 is an acknowledged TD-learning reinterpretation; the core taxonomy itself remains an independent survey contribution.

  1. renaming known result [Section 6.1, Equation 4]
    "this paper proposes a SHAP-weighted explainability metric E (Equation 4) ... E captures the degree to which influential features are also interpretable. A higher E indicates that features with greater predictive influence are also more explainable."

    E is defined as the weighted L1 norm of SHAP values: E = Σ |ω_i * SHAP_i|. The formula contains no term that measures interpretability, auditability, or the degree to which a feature is explainable; the sentence 'higher E indicates ... more explainable' is an assertion attached to the definition, not a consequence of the equation. With ω_i read as 'relative importance weight', E is merely an attribution-magnitude score that increases whenever high-SHAP features are present. With ω_i read as an interpretability weight, the paper provides no definition or construction for such a weight and contradicts the meaning assigned in the same paragraph. In either reading, the paper's explainability metric reduces by construction to a weighted sum of SHAP magnitudes renamed as 'explainability'.

  2. renaming known result [Section 7.2, Equation 5]
    "α_{t+1} = α_t + η(R_t − α_t) ... The update rule, though mathematically similar to temporal-difference learning, is reinterpreted here within the context of alpha generation—treating α_t as a dynamic estimate of strategic performance rather than as a traditional value function."

    Equation 5 is exactly the temporal-difference (TD) error-correction rule, α ← α + η(r − α), with α renamed from a value estimate to an 'estimate of alpha' and R_t renamed as realized return. The paper explicitly acknowledges the mathematical identity to TD learning, yet still presents the equation as a 'typical update mechanism' for RL-LLM alpha agents. No RL-LLM mechanism enters the equation; the contribution is a variable renaming of a known update rule. Because the paper itself labels it a reinterpretation, this is a transparent renaming rather than a hidden derivation.

full rationale

The paper is primarily a taxonomy and survey, not an empirical derivation, so most of its content is not circular: the five-stage classification, the historical review, and the system-level framing stand as independent organizational work. The main circularity burden falls on the proposed quantitative governance metrics. Equation 4 is presented as a SHAP-weighted explainability metric, but by construction it is only a weighted absolute-SHAP sum; the claim that a higher value means features are more explainable is an interpretive label, not a result of the equation. Equation 5 is a direct reuse of the temporal-difference update rule; the paper admits this, so it is an acknowledged renaming rather than a concealed derivation. The Trust Score in Equation 3 is an operational definition of trust as a weighted sum of named dimensions; while it lacks external validation, it is not circular in the same reduction sense. The frequent self-citations to the author's own book [30] support broad governance claims but are not load-bearing for the taxonomy or for any fitted prediction, so they do not raise the score substantially. Overall, the central taxonomic claim retains independent content, but the explainability metric's stated meaning is forced by its own definition, giving a partial circularity score of 4.

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

The paper introduces no new physical or conceptual entities; the Trust Score is a composite metric, not an entity. The main extra assumptions are the linear ordering of the taxonomy and the measurability of trust components, both of which are asserted rather than derived.

free parameters (4)
  • omega_1, omega_2, omega_3, omega_4 (Trust Score weights)
    Introduced in Eq. 3 as 'tunable' weights with no default values, calibration procedure, or external validation; the claimed trust measurement depends on them.
  • W_t, W_s, W_g, b (fusion weights and bias in Eq. 2)
    Learnable parameters of a generic multimodal alpha-score function; no training data, fitted values, or justification of separability are provided.
  • w_i (SHAP importance weights in Eq. 4)
    Relative importance weights for each feature in the explainability metric; no assignment rule or values are given.
  • eta (learning rate in Eq. 5)
    Reinforcement learning update step size in Eq. 5; no value or tuning protocol is provided.
assumptions (4)
  • ad hoc to paper Alpha can be meaningfully decomposed by the five-stage linear taxonomy from manual to LLM agents.
    Section 5 asserts the five-stage maturity ordering without systematic evidence or a comparison against alternative taxonomies; real systems often combine stages.
  • ad hoc to paper The components of Trust Score (Attribution, Stability, Factuality, Alignment) are measurable and linearly composable.
    Section 6.1 Eq. 3 assumes these four dimensions can be scored, normalized, and combined with weights to produce a valid trust measure; no operational definitions are supplied.
  • domain assumption SHAP values provide a sound basis for a compliance-oriented explainability metric.
    Section 6.1 Eq. 4 relies on the standard assumption that SHAP attributions approximate feature influence; the paper does not address known limitations of SHAP for correlated financial features.
  • domain assumption LLM agents with tool use and memory can act as financial co-processors.
    Sections 4.2 and 5 Stage 5 treat agentic capabilities as established based on cited frameworks such as LangChain, AutoGPT, and FinAgent; the paper does not measure reliability in live markets.

how reviews work

0 comments
Cite this review

Pith. "Pith review of The Evolution of Alpha in Finance Harnessing Human Insight and LLM Agents." pith.science (2026). https://pith.science/paper/WFNRSIRO

@misc{pith2026250514727,
  author       = {Pith},
  title        = {Pith review of: The Evolution of Alpha in Finance Harnessing Human Insight and LLM Agents},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WFNRSIRO}},
  note         = {Machine review of arXiv:2505.14727}
}
read the original abstract

The pursuit of alpha returns that exceed market benchmarks has undergone a profound transformation, evolving from intuition-driven investing to autonomous, AI powered systems. This paper introduces a comprehensive five stage taxonomy that traces this progression across manual strategies, statistical models, classical machine learning, deep learning, and agentic architectures powered by large language models (LLMs). Unlike prior surveys focused narrowly on modeling techniques, this review adopts a system level lens, integrating advances in representation learning, multimodal data fusion, and tool augmented LLM agents. The strategic shift from static predictors to contextaware financial agents capable of real time reasoning, scenario simulation, and cross modal decision making is emphasized. Key challenges in interpretability, data fragility, governance, and regulatory compliance areas critical to production deployment are examined. The proposed taxonomy offers a unified framework for evaluating maturity, aligning infrastructure, and guiding the responsible development of next generation alpha systems.

Figures

Figures reproduced from arXiv: 2505.14727 by the authors.

Figure 2
Figure 2. Fig2 [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Workflow of an LLM-Based Agent in Financial Contexts These capabilities allow LLMs to act not merely as forecasters, but as embedded financial co-processors—supporting functions across trading, risk monitoring, compli￾ance, and research. By consolidating perception, reasoning, and execution within a sin￾gle architecture, agentic LLMs reduce operational fragmentation and augment human decision-making with continuous,… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

72 extracted references · 64 canonical work pages

  1. [1]

    J Finance 19(3):425–442

    Sharpe WF (1964) Capital asset prices: A theory of market equilibrium under conditions of risk. J Finance 19(3):425–442

    work page 1964

  2. [2]

    McGraw-Hill, New York

    Graham B, Dodd DL (1934) Security analysis. McGraw-Hill, New York

    work page 1934

  3. [3]

    Hermes: Columbia Busi- ness School Magazine

    Buffett W (1984) The superinvestors of Graham -and-Doddsville. Hermes: Columbia Busi- ness School Magazine

    work page 1984

  4. [4]

    New York Institute of Fi- nance, New York

    Murphy JJ (1999) Technical analysis of the financial markets. New York Institute of Fi- nance, New York

    work page 1999

  5. [5]

    J Finance 55(4):1705–1765

    Lo AW, Mamaysky H, Wang J (2000) Foundations of technical analysis: computational algorithms, statistical inference, and empirical implementation. J Finance 55(4):1705–1765

    work page 2000

  6. [6]

    J Finance 23(2):389–416

    Jensen MC (1968) The performance of mutual funds in the period 1945 –1964. J Finance 23(2):389–416

    work page 1968

  7. [7]

    J Econ Theory 13(3):341–360

    Ross SA (1976) The arbitrage theory of capital asset pricing. J Econ Theory 13(3):341–360

    work page 1976

  8. [8]

    J Financ Econ 33(1):3–56

    Fama EF, French KR (1993) Common risk factors in the returns on stocks and bonds. J Financ Econ 33(1):3–56

    work page 1993

Show all 72 references
  1. [9]

    Princeton University Press, Princeton

    Lo AW, MacKinlay AC (1999) A non-random walk down Wall Street. Princeton University Press, Princeton

    1999

  2. [10]

    Wiley, New York

    Chan E, Chan EP (2013) Algorithmic trading: winning strategies and their rationale. Wiley, New York

    2013

  3. [11]

    HFR Industry Reports (2023) Hedge Fund Research, Inc

    2023

  4. [12]

    J Financ Data Sci 1(2):64–74 31

    Arnott R, Harvey CR, Markowitz HM (2019) A backtesting protocol in the era of machine learning. J Financ Data Sci 1(2):64–74 31

    2019

  5. [13]

    Rev Financ Stud 33(5):2223–2273

    Gu S, Kelly B, Xiu D (2020) Empirical asset pricing via machine learning. Rev Financ Stud 33(5):2223–2273

    2020

  6. [14]

    Eur J Oper Res 259(2):689–702

    Krauss C, Do XA, Huck N (2017) Deep neural networks, gradient -boosted trees, random forests: statistical arbitrage on the S&P 500. Eur J Oper Res 259(2):689–702

    2017

  7. [15]

    Rev Financ Stud 34(9):4362–4419

    Avramov D, Cheng S, Metzker Z (2021) Machine learning, clustering, and asset pricing. Rev Financ Stud 34(9):4362–4419

    2021

  8. [16]

    arXiv:2303.08774

    OpenAI (2023) GPT-4 technical report. arXiv:2303.08774

    2023 arXiv

  9. [17]

    arXiv:2307.09288

    Meta AI (2023) LLaMA 2: open foundation and chat models. arXiv:2307.09288

    2023 arXiv

  10. [18]

    arXiv:2306.12573

    Yang Y, Ma Z, Liu W et al (2023) FinGPT: large language model for finance with reinforce- ment learning. arXiv:2306.12573

    2023 arXiv

  11. [19]

    arXiv:2112.09332

    Chen M, Tworek J, Jun H et al (2021) Evaluating LLMs as financial agents. arXiv:2112.09332

    2021 arXiv

  12. [20]

    arXiv:2303.17564

    Wu J, Hu C, Zhang Y et al (2023) BloombergGPT: a large language model for finance. arXiv:2303.17564

    2023 arXiv

  13. [21]

    IEEE Trans Comput Soc Syst 6(3):456–465

    Deng Y, Bao F, Kong Y, Ren Z, Dai Q (2019) Deep learning for financial sentiment analysis on Twitter and stock prediction. IEEE Trans Comput Soc Syst 6(3):456–465

    2019

  14. [22]

    arXiv:2006.00695

    Araci D (2019) FinBERT: financial sentiment analysis with pre -trained language models. arXiv:2006.00695

    2019 arXiv

  15. [23]

    arXiv:2308.11458

    Mialon G, Villalobos J, Luccioni A et al (2023) AgentBench: evaluating LLMs as agents. arXiv:2308.11458

    2023

  16. [24]

    arXiv:2311.09187

    Xu K, Liu P, Zhang X et al (2023) AutoFinAgent: towards autonomous financial decision - making agents using LLMs. arXiv:2311.09187

    2023 arXiv

  17. [25]

    ACM/IMS J Data Sci

    Cao B, Wang S, Lin X et al (2025) From deep learning to LLMs: a survey of AI in quanti- tative investment. ACM/IMS J Data Sci

    2025

  18. [26]

    Quant Finance 21(11):1773–1792

    Hu Z, Zhao Y, Huang JZ (2021) Multimodal learning in financial applications: a review. Quant Finance 21(11):1773–1792

    2021

  19. [27]

    MIT Press, Cambridge

    Goodfellow I, Bengio Y, Courville A (2016) Deep learning. MIT Press, Cambridge

    2016

  20. [28]

    Brussels

    European Commission (2021) Proposal for a regulation laying down harmonised rules on artificial intelligence (AI Act). Brussels

    2021

  21. [29]

    SEC (2023) Staff bulletin: standards of conduct for investment advisers and broker -dealers regarding AI technologies

    2023

  22. [30]

    Wiley, New York

    Islam MR (2024) Generative AI, cybersecurity, and ethics. Wiley, New York

    2024

  23. [31]

    Wiley, New York

    Schwager JD (1989) Market wizards: interviews with top traders. Wiley, New York

    1989

  24. [32]

    In: Proc 22nd ACM SIGKDD Int Conf Knowl Discov Data Min, pp 785–794

    Chen T, Guestrin C (2016) XGBoost: a scalable tree boosting system. In: Proc 22nd ACM SIGKDD Int Conf Knowl Discov Data Min, pp 785–794

    2016

  25. [33]

    J Bank Finance 34(11):2767–2787

    Khandani AE, Kim AJ, Lo AW (2010) Consumer credit -risk models via machine-learning algorithms. J Bank Finance 34(11):2767–2787

    2010

  26. [34]

    Mach Learn 45(1):5–32

    Breiman L (2001) Random forests. Mach Learn 45(1):5–32

    2001

  27. [35]

    Appl Econ Lett 27(17):1401–1407

    Hoang T, McNelis PD (2020) Forecasting cross-sectional stock returns with gradient boost- ing machine. Appl Econ Lett 27(17):1401–1407

    2020

  28. [36]

    Appl Stoch Models Bus Ind 33(1):3–12

    Heaton JB, Polson NG, Witte JH (2017) Deep learning for finance: deep portfolios. Appl Stoch Models Bus Ind 33(1):3–12

    2017

  29. [37]

    Int Jt Conf Neural Netw (IJCNN)

    Nelson DM, Pereira ACM, de Oliveira RA (2017) Stock market’s price movement predic- tion with LSTM neural networks. Int Jt Conf Neural Netw (IJCNN)

    2017

  30. [38]

    J Comput Finance 20(4):39–69

    Bailey DH, Borwein JM, López de Prado M, Zhu QJ (2014) The probability of backtest overfitting. J Comput Finance 20(4):39–69

    2014

  31. [39]

    Quant Finance 18(12):2043–2055 32

    Huang T, Liu Y, Yang Y (2018) A new machine learning framework for forecasting stock return classification. Quant Finance 18(12):2043–2055 32

    2018

  32. [40]

    In: Adv Neural Inf Process Syst (NeurIPS)

    Lundberg SM, Lee S-I (2017) A unified approach to interpreting model predictions. In: Adv Neural Inf Process Syst (NeurIPS)

    2017

  33. [41]

    Springer, New York

    Vapnik VN (1995) The nature of statistical learning theory. Springer, New York

    1995

  34. [42]

    (2021) Deep Learning for Time -Series Alpha Generation

    Zhang, Y., Zhang, J., & Qin, Z. (2021) Deep Learning for Time -Series Alpha Generation. Journal of Financial Data Science

    2021

  35. [43]

    Eur J Oper Res 270(2):654–669

    Fischer T, Krauss C (2018) Deep learning with long short-term memory networks for finan- cial market predictions. Eur J Oper Res 270(2):654–669

    2018

  36. [44]

    IEEE Trans Knowl Data Eng

    Wang Y, Zhang C, Shi Y (2021) Graph -based financial forecasting with multimodal fea- tures. IEEE Trans Knowl Data Eng

    2021

  37. [45]

    arXiv:1702.08608

    Doshi-Velez F, Kim B (2017) Towards a rigorous science of interpretable machine learning. arXiv:1702.08608

    2017 arXiv

  38. [46]

    In: Adv Neural Inf Process Syst (NeurIPS)

    Xu Y, Cohen W, Wallach H (2019) A neural multimodal model for multivariate time series forecasting. In: Adv Neural Inf Process Syst (NeurIPS)

    2019

  39. [47]

    Expert Syst Appl 42(1):613–632

    Nassirtoussi AK, Aghabozorgi S, Wah TY, Ngo DCL (2015) Text mining for market pre- diction: a systematic review. Expert Syst Appl 42(1):613–632

    2015

  40. [48]

    J Financ Data Sci 4(2):22–37

    Demskoi D, Grishina M (2022) Using transformer models to quantify tone in earnings calls. J Financ Data Sci 4(2):22–37

    2022

  41. [49]

    arXiv:2305.11764

    He J, Fan K, Zhao W et al (2023) A survey of hallucination in large language models. arXiv:2305.11764

    2023 arXiv

  42. [50]

    Int J Forecast 37(4):1748–1764

    Lim B, Arık SÖ, Loeff N, Pfister T (2021) Temporal fusion transformers for interpretable multivariate time series forecasting. Int J Forecast 37(4):1748–1764

    2021

  43. [51]

    PLOS ONE 12(7):e0180944

    Bao W, Yue J, Rao Y (2017) A deep learning framework for financial time series using stacked autoencoders and LSTM. PLOS ONE 12(7):e0180944

    2017

  44. [52]

    Springer, Heidelberg

    Samek W, Montavon G, Vedaldi A, Hansen LK, Müller KR (2019) Explainable AI: inter- preting, explaining and visualizing deep learning. Springer, Heidelberg

    2019

  45. [53]

    Commun ACM 61(10):36–43

    Lipton ZC (2018) The mythos of model interpretability. Commun ACM 61(10):36–43

    2018

  46. [54]

    Wiley, New York

    López de Prado M (2018) Advances in financial machine learning. Wiley, New York

    2018

  47. [55]

    In: Proc 22nd ACM SIGKDD Int Conf Knowl Discov Data Min

    Ribeiro MT, Singh S, Guestrin C (2016) Why should I trust you? Explaining the predictions of any classifier. In: Proc 22nd ACM SIGKDD Int Conf Knowl Discov Data Min

    2016

  48. [56]

    Big Data 5(3):246–255

    Varshney KR, Alemzadeh H (2017) On the safety of machine learning: cyber-physical sys- tems, decision sciences, and data products. Big Data 5(3):246–255

    2017

  49. [57]

    ACM Trans Inf Syst 37(2):1–30

    Feng F, He X, Wang X, Luo C, Liu Y, Chua TS (2019) Temporal relational ranking for stock prediction. ACM Trans Inf Syst 37(2):1–30

    2019

  50. [58]

    Adv Neural Inf Process Syst (NeurIPS) 33:1877–1901

    Brown T, Mann B, Ryder N et al (2020) Language models are few-shot learners. Adv Neural Inf Process Syst (NeurIPS) 33:1877–1901

    2020

  51. [59]

    arXiv:2108.07258

    Bommasani R, Hudson D, Liang P (2021) On the opportunities and risks of foundation mod- els. arXiv:2108.07258

    2021 arXiv

  52. [60]

    J Perform Meas 16(3):23–33

    Menchero J, Hu J (2012) Portfolio risk attribution with factor models. J Perform Meas 16(3):23–33

    2012

  53. [61]

    ACM Comput Surv 55(12):1–38

    Ji Z, Lee N, Frieske R et al (2023) Survey of hallucination in natural language generation. ACM Comput Surv 55(12):1–38

    2023

  54. [62]

    arXiv:2102.07350

    Perez E, Ribeiro MT, Kiela D (2022) True few -shot learning with language models: the inadequacy of prompt-based learning. arXiv:2102.07350

    2022 arXiv

  55. [63]

    Trans Assoc Comput Linguist

    Schick T, Schütze H (2021) Generating agentic behavior in large language models. Trans Assoc Comput Linguist

    2021

  56. [64]

    Commun ACM 64(12):70–78

    Patterson D, Gonzalez J, Le Q et al (2021) Carbon emissions and large neural networks. Commun ACM 64(12):70–78

    2021

  57. [65]

    Adv Neural Inf Process Syst (NeurIPS) 35:13038–13051 33

    Lin Z, Hilton J, Evans O (2022) TruthfulQA: measuring how models mimic human false- hoods. Adv Neural Inf Process Syst (NeurIPS) 35:13038–13051 33

    2022

  58. [66]

    Sci- ence 349(6245):255–260

    Jordan MI, Mitchell TM (2015) Machine learning: trends, perspectives, and prospects. Sci- ence 349(6245):255–260

    2015

  59. [67]

    J Financ Econ 21(2):255–289

    Connor G, Korajczyk RA (1988) Risk and return in an equilibrium APT: application of a new test methodology. J Financ Econ 21(2):255–289

    1988

  60. [68]

    Knowl Eng Rev 10(2):115–152

    Wooldridge M, Jennings NR (1995) Intelligent agents: theory and practice. Knowl Eng Rev 10(2):115–152

    1995

  61. [69]

    Stat Sci 17(3):235–255

    Bolton RJ, Hand DJ (2002) Statistical fraud detection: a review. Stat Sci 17(3):235–255

    2002

  62. [70]

    J Finance 7(1):77–91

    Markowitz H (1952) Portfolio selection. J Finance 7(1):77–91

    1952

  63. [71]

    In: Proc Int Conf Mach Learn (ICML)

    Zhang Y, Levine S, Abbeel P (2023) Decision transformer: reinforcement learning via se- quence modeling. In: Proc Int Conf Mach Learn (ICML)

    2023

  64. [72]

    ACM Comput Surv 54(8):1–36

    He X, Zhao K, Chu X (2021) AutoML: a survey of the state-of-the-art. ACM Comput Surv 54(8):1–36

    2021

Pith tools

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