Pith. sign in

REVIEW 3 major objections 6 minor 65 references

TradingMoE: Routing the Right Experts in Evolving Markets

T0 review · 3 major / 6 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read TradingMoE claims that a query–key router with sparse expert replacement lifts cumulative trading returns by roughly 30 percentage points over the best baselines on stocks and crypto.

desk verdict Genuinely new MoE routing mechanism, but the headline stock gain is not leakage-controlled; the Crypto control and the method itself justify a serious look. read the letter →

arxiv 2608.11785 v1 pith:5REMI7MX submitted 2026-08-12 cs.LG

Chang Zhou , Xingtong Yu , Minbin Huang , Zhennan Wu , Yuan Fang , Hong Cheng , Xinming Zhang This is my paper · ORCID
classification cs.LG
keywords largelanguagemodelsai4tradingmixture-of-expertsexpertroutingquery-keyroutercounterfactualcreditlow-rankstructurepapertrading
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

TradingMoE tries to establish that the bottleneck in LLM-based trading is not the language model itself but the router that decides which specialized experts process each token of the trading decision. It reports that conventional internal MoE routers assign scores that barely track how much an expert actually improves the decision (correlation of −0.015), and that expert usefulness across tokens lives in a compact, approximately low-rank structure. On that basis, the paper freezes a pretrained dense LLM, adds lightweight residual experts, and routes them with a query–key matcher that compares a per-token demand query with learnable expert keys, plus a mechanism that samples inactive experts and lets them replace the weakest active expert when market context shifts. Against 22 baselines, TradingMoE reports cumulative returns of 49.08% on the stock test set and 73.79% on a held-out crypto year — 30.89 and 30.7 percentage points above the strongest baselines, and 37.81 and 80.17 percentage points above buy-and-hold. If these margins survive forward-only deployment, the method offers a recipe for specializing frozen LLMs to evolving markets without retraining the backbone.

What carries the argument

The load-bearing object is the counterfactual expert credit, a first-order estimate of how replacing an active expert with a candidate would change the trading-decision loss. For token $t$, layer $\ell$, and expert $j$ it is $C^{\ell}_{t,j} = -\langle \partial \mathcal{L}_{\mathrm{task}}/\partial \mathbf{h}^{\ell}_t,\, e^{\ell}_j(\mathbf{h}^{\ell}_t) \rangle$; relative to the weakest active expert it becomes $\Delta C^{\ell}_{t,i} = \alpha^{\ell}_{t,\mathrm{low}}(C^{\ell}_{t,i} - C^{\ell}_{t,\mathrm{low}})$, which preserves within-token expert ordering. This credit object plays three roles: its measured low-rank structure motivates parameterizing the router's score matrix as $\mathbf{S}^{\ell} = \mathbf{Q}^{\ell}(\mathbf{U}^{\ell})^{\top}$ with $\mathrm{rank}(\mathbf{S}^{\ell}) \le d_q$ (the paper uses $d_q = 16$ and 64 experts); the query–key router realizes that factorization; and the sparse selection update injects a signal $\boldsymbol{\xi}^{\ell}_t$ that cancels in the forward pass but, during backpropagation, raises or lowers each sampled inactive expert's routing margin exactly when first-order credit predicts replacement would help or hurt. Two propositions guarantee the mechanism is well-behaved: the sampled margin gradient is an unbiased estimate of the average over all inactive experts, and the induced router update is consistent with the local loss reduction of expert replacement.

What would settle it

The decisive check is a leakage-controlled rerun of the Stock backtest: train TradingMoE with its protocol fixed but on a backbone whose pretraining corpus provably ends before the 2023-06-13 test start, and compare cumulative return against LightGBM over the same 138 trading days; if the margin falls far below the reported 30.89 percentage points, memorization, not routing, carried the headline result. A routing-specific check is to apply the paper's own diagnostic to TradingMoE's trained router: if the query–key margins show near-zero or negative correlation with measured replacement gains on held-out decision tokens, the premise that counterfactual credit improves expert selection would be contradicted.

Watch

Extended reading notes

Core claim

The central claim is that expert routing for trading should be driven by counterfactual expert credit — an estimate of how much swapping one expert for another would reduce the trading-decision loss — rather than by naive router scores trained only through the task loss. Two empirical findings ground this claim. First, native router scores have essentially zero correlation with measured replacement gains (−0.015), and 66.76% of decision tokens have at least one unselected expert that would have improved the decision. Second, token–expert credit matrices are approximately low-rank, with rank-16 reconstructions retaining 74.2% and 77.9% of credit energy on the OLMoE and DeepSeek-V2-Lite backbones. TradingMoE implements the finding as a query–key router, in which a low-dimensional per-token query encodes the expertise demanded under the current market context and is matched against learnable expert keys, so the score matrix $\mathbf{S}^{\ell} = \mathbf{Q}^{\ell}(\mathbf{U}^{\ell})^{\top}$ has rank at most the query dimension. A sparse expert selection update then samples a few inactive experts per decision token, compares each against the lowest-scored active expert, and adjusts routing margins in the same backward pass; the paper proves the sampled update is unbiased over inactive experts and directionally consistent with first-order loss reduction from replacement. On the reported benchmarks this yields 49.08% cumulative return on the stock test set and 73.79% on the crypto year, beating the strongest baseline in each market by 30.89 and 30.7 percentage points.

Load-bearing premise

The headline margins assume that the frozen language model's pretraining did not effectively give it knowledge of the stock and crypto evaluation periods, so the reported gains come from the routing method rather than from memorized market outcomes.

Editorial extensions

If this is right

  • Frozen dense LLMs can be specialized to trading cheaply: only small residual experts, expert keys, and query heads are trained, while the pretrained backbone stays untouched and the added computation stays sparse (top-$k$ of 4 among 64 experts per layer).
  • Router design becomes the primary lever for LLM trading performance: the diagnostics show naive router scores carry almost no information about expert contribution, and the ablation attributes most of the gain to the query–key credit-based routing rather than to the selection update alone.
  • The same model, trained only on U.S. stocks, transfers to a different asset class: it ranks first on the held-out crypto year and stays positive in forward-only paper trading, suggesting the routing mechanism rather than market-specific memorization drives the edge.
  • Market-regime shifts are handled by construction: inactive experts are periodically sampled as challengers and can replace the weakest active expert, so the route can change as the market context encoded in the token queries changes.
  • The stock advantage is not a single-run artifact: across five seeds the margin over LightGBM is significant under a robust Sharpe-difference test (two-sided $p = 0.018$) and a HAC-adjusted mean test ($p = 0.007$).

Reading between the lines

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

  • The low-rank credit structure is measured only on two pretrained MoE backbones; if it is a general property of expert usefulness in non-stationary decision tasks, the query–key router would transfer to other LLM decision problems where the right specialization shifts with context — my inference, not demonstrated in the paper.
  • The Stock test window falls inside the training corpus of any 2024-or-later backbone, so the cleanest confirmation of the 30.89-percentage-point stock margin would be a re-run with a backbone whose pretraining provably ends before mid-2023 — a control the leakage-controlled section does not include.
  • Because the selection update samples only $m = 2$ of the roughly 60 inactive experts per token, regime adaptation is limited by sampling noise; making the sampling credit-weighted or increasing $m$ could sharpen responsiveness to regime shifts at modest extra compute.
  • The reported cost sensitivity implies the method's edge survives roughly twice the one-way transaction-cost rate that breaks the best baselines (about 28 bps vs. 13–14 bps), so the practical deployment window is wider than for the alternatives even though daily rebalancing eventually becomes unprofitable at extreme costs.
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 proposes TradingMoE, a sparse mixture-of-experts layer for frozen LLMs used in daily trading-decision generation. It introduces a query-key router that computes low-dimensional token queries and matches them with learnable expert keys, and a sparse expert-selection update that samples inactive experts, compares them with the lowest-scored active expert, and injects a detached routing-margin signal in the same backward pass. Two theoretical propositions show that inactive-expert sampling yields an unbiased gradient estimate and that the router update is consistent with first-order replacement benefit. Experiments on a Stock benchmark (test window 2023-06-13 to 2023-12-31) and a Crypto benchmark (2025) report cumulative returns of 49.08% and 73.79%, outperforming the best baselines by 30.89 and 30.78 percentage points. Section 5.4 additionally reports a forward-only Stock paper-trading run and a Qwen2.5-7B Crypto control.

Significance. If the results are trusted, the paper makes a useful contribution to LLM-based trading: it documents a concrete failure of native MoE routers (Pearson correlation -0.015 with replacement gains), identifies a low-rank structure in token-expert credit, and proposes a training mechanism that is computationally sparse. Strengths include the authors' explicit acknowledgment of temporal-leakage risk, a five-seed statistical robustness protocol (Appendix B.5), transaction-cost sensitivity analysis (Appendix E.3), and the fact that Propositions 1 and 2 are consistency/identifiability statements about the proposed gradient update rather than circular derivations of the empirical results. The controlled Crypto experiment in Table 5 suggests a genuine but possibly smaller advantage. The main unresolved point is that the headline Stock margin currently lacks a leakage-free quantitative estimate.

major comments (3)
  1. [§5.4, Table 5, Figure 4] The headline Stock result is not leakage-controlled: the backbone Qwen3.5-9B was released in 2026-03, after the Stock test window (2023-06-13 to 2023-12-31), so its pretraining corpus may contain information about the evaluation period. The only forward-only Stock evidence is a three-month paper-trading run shown in Figure 4, which reports no numerical metrics and no baseline margin. Consequently the abstract's 30.89pp improvement over LightGBM rests entirely on a post-dating backbone, and the claim is not cleanly established. The Crypto control in Table 5 is valid for Crypto and reduces the margin from 30.78pp to 25.37pp, but it cannot control for the Stock window because Qwen2.5-7B also postdates 2023. Please provide a Stock-specific leakage control (e.g., a backbone whose pretraining corpus provably excludes the Stock test window, or a comparable differential test) or clearly state that the headline Stock margin is measured only in the uncontrolled setting.
  2. [§5.4, final paragraph] The sentence 'the potential temporal overlap has only a limited impact' is supported only by a Crypto comparison between TradingMoE(Qwen2.5-7B) and TradingMoE(Qwen3.5-9B). That comparison says nothing about the Stock benchmark, where the backbone postdates the test window. The subsequent conclusion that 'the overall advantage primarily comes from the proposed method rather than temporal leakage' is therefore stronger than the evidence supports for Stock.
  3. [§5.4, Figure 4] The forward-only paper-trading experiment is reported only as a figure without cumulative-return values, Sharpe ratios, or a table of baseline margins; moreover, the DeepSeek curve switches from DeepSeek-V3.2 to DeepSeek-V4-Pro mid-period (Appendix D), which may confound that particular comparison. Without numeric results, the statement that the advantage persists under forward-only deployment is not quantitatively verifiable.
minor comments (6)
  1. [§5.2] In the first observation about Table 2, 'it it obtains' should read 'it obtains'.
  2. [Table 2] In the FinAgent row, '1.5624.57%' appears to be two values run together; please separate the Sharpe ratio and maximum drawdown for readability.
  3. [§5.4] The date '2026-6-30' should be written as '2026-06-30' for consistency with the other dates in the paper.
  4. [§5.3] The phrase 'spare expert selection update' should be 'sparse expert selection update'.
  5. [Appendix A.4] The authors correctly state that 'SVD fitting and comparison use the same matrix, so this diagnostic evaluates representational capacity rather than held-out prediction.' Because Table 1 is used as motivation for the low-rank router, this caveat should also appear near Table 1 in the main text so readers do not overinterpret the 98% better-than-native route rate as a predictive result.
  6. [Figure 4] Add axis labels and units to Figure 4, and report the terminal cumulative returns of all displayed curves in the text or caption so the forward-only result can be assessed numerically.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the router and expert-update training signals are derived from the external task loss by backpropagation, and the one self-referential diagnostic (App. A.4) is disclosed and does not drive the headline results.

full rationale

The derivation chain of TradingMoE is not circular. The router scores are computed from token representations by Eqs. 13–15, and the sparse expert selection update injects a zero-forward value (Eq. 18) whose gradient is derived, via the chain rule, from the task loss (Eq. 12). Propositions 1 and 2 are consistency statements about sampling and gradient direction, not fitted quantities relabeled as predictions. The claimed 30.89% and 30.7% improvements are measured on held-out chronological test windows against 22 baselines, so the central empirical claim is externally falsifiable rather than being equivalent to an input. The one self-referential element is the low-rank diagnostic in App. A.4, which fits SVD reconstructions on the same token-expert credit matrix it evaluates; however, the paper explicitly discloses that 'SVD fitting and comparison use the same matrix, so this diagnostic evaluates representational capacity rather than held-out prediction,' and this diagnostic does not generate the reported returns. The Section 5.4 leakage concern is a temporal-overlap validity risk, not a circularity: the Qwen2.5-7B Crypto control and the forward-only paper-trading run are attempts to address it, and the paper explicitly concedes the risk. No load-bearing argument reduces to a self-citation; the references to first-order credit (Michel et al., Molchanov et al.) are standard external sources. Overall, the derivation is self-contained, and the empirical claims are not forced by construction.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The central claim rests on standard math (Taylor expansion, unbiased sampling) and on domain assumptions about the simulation and the backbone's pretraining window. No new entities are introduced. The main free parameters are router-capacity hyperparameters selected on the Stock validation set.

free parameters (5)
  • query dimension d_q = 16
    Dimensionality of the low-rank routing query and keys. Selected by hyperparameter search on Stock validation (App. E.1); performance is sensitive to joint configuration with E.
  • number of experts E = 64
    Number of LoRA experts per routed layer. Validated jointly with d_q (App. E.1).
  • Top-k and sampled challengers (k, m) = k=4, m=2
    Sparsity level and number of inactive experts sampled per token in the selection update; set by validation.
  • selection-update strength lambda_sel = 0.05
    Scales the injected router-update signal in Eq. 17; hyperparameter.
  • LoRA rank r_L and alpha_L = 12 and 24
    Capacity of the residual experts; standard LoRA hyperparameters.
assumptions (4)
  • standard math First-order Taylor expansion and uniform sampling without replacement give an unbiased, replacement-consistent router update (Propositions 1 and 2).
    The proofs in Appendices A.1 and A.2 rely on standard linearity of expectation and smoothness of the task loss; these are background math assumptions.
  • domain assumption The portfolio simulator faithfully represents the deployed strategy: open-to-close fills, 5 bps one-way cost, unit gross exposure, no slippage, no borrow fees.
    Appendix B.3 defines the simulation; its realism is untested and affects all reported returns.
  • domain assumption The frozen LLM backbone's pretraining corpus does not materially overlap with the evaluation periods.
    Section 5.4 explicitly flags this risk for LLM financial backtests and provides partial controls (Qwen2.5-7B for Crypto, 2026 paper trading for Stock), but the primary Tables 2 and 3 use Qwen3.5-9B released 2026-03 for earlier periods.
  • domain assumption Ground-truth decisions built from largest same-day absolute open-to-close returns (App. B.1) are a learnable, stationary training target.
    The labels depend on ex-post daily moves; the model is trained to predict these magnitudes from previous information. If the signal is not learnable out-of-sample, the performance advantage would vanish.

how reviews work

0 comments
Cite this review

Pith. "Pith review of TradingMoE: Routing the Right Experts in Evolving Markets." pith.science (2026). https://pith.science/paper/5REMI7MX

@misc{pith2026260811785,
  author       = {Pith},
  title        = {Pith review of: TradingMoE: Routing the Right Experts in Evolving Markets},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5REMI7MX}},
  note         = {Machine review of arXiv:2608.11785}
}
read the original abstract

Large language models (LLMs) have shown strong potential for financial analysis and trading, but direct trading remains challenging because the predictive capabilities required can vary across assets, decision fields, and market conditions. Existing LLM-based trading systems either coordinate human-defined external experts or adopt conventional internal Mixture-of-Experts (MoE) routers that do not directly evaluate how individual experts contribute to trading decisions. Moreover, these routers receive no direct signal indicating when an inactive expert has become more suitable as market conditions change. We find that native router scores poorly reflect how much individual experts improve trading decisions, frequently leaving better alternatives unselected. We further reveal that token-specific expert usefulness exhibits a compact low-dimensional structure. Based on these findings, we propose TradingMoE, a trading-oriented sparse MoE that augments a frozen dense LLM with lightweight residual experts. We introduce a Query-Key router that represents the expertise required by each token under the current market context as a low-dimensional query and matches it with learnable expert keys. We further propose a sparse expert selection update mechanism that samples a few inactive experts during training and estimates whether they should replace the weakest expert in the current Top-k route. This mechanism enables the router to update expert selection as market conditions change while preserving sparse computation. Experiments against 22 baselines on stock and cryptocurrency markets show that TradingMoE improves cumulative return over the best-performing baselines by 30.89% and 30.7%, respectively. Rolling paper-trading experiments further demonstrate that its advantage persists under forward-only deployment.

Figures

Figures reproduced from arXiv: 2608.11785 by the authors.

Figure 1
Figure 1. Comparison of expert routing methods. Existing LLM-based trading systems commonly realize special￾ization through external expert routing, as shown in [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Overview of TradingMoE. the key of 𝑒 ℓ 𝑙𝑜𝑤 (·). A positive Δ𝑠 ℓ 𝑡,𝑖 indicates that the sampled inac￾tive expert receives a higher routing score than the lowest-scored active expert. Same-step router update. We combine the detached routing mar￾gins with the corresponding expert-output differences: 𝝃 ℓ 𝑡 = 𝜆sel 𝑚 ∑︁ 𝑖∈Iˆℓ 𝑡 Δ𝑠 ℓ 𝑡,𝑖stopgrad(𝛼 ℓ 𝑡,low h 𝑒 ℓ 𝑖  h ℓ 𝑡  − 𝑒 ℓ 𝑙𝑜𝑤  h ℓ 𝑡  i), (17) where 𝜆sel > 0 contro… view at source ↗
Figure 3
Figure 3. Cumulative return curves on Stock and Crypto. [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Cumulative return in the paper-trading. 6 Conclusions In this paper, we proposed TradingMoE, a trading-oriented sparse internal expert routing framework for direct trading-decision gener￾ation with LLMs. Motivated by the compact structure of counterfac￾tual expert cont…
Figure 5
Figure 5. Figure 5: Impact of expert count and query dimension. [PITH_FULL_IMAGE:figures/full_fig_p017_5.png]
Figure 6
Figure 6. Figure 6: Cumulative return on Stock under varying one-way [PITH_FULL_IMAGE:figures/full_fig_p017_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

65 extracted references · 40 canonical work pages

  1. [1]

    Andrew Ang and Allan Timmermann. 2012. Regime changes and financial markets.Annu. Rev. Financ. Econ.4, 1 (2012), 313–337

    work page 2012

  2. [2]

    John S Bridle. 1990. Probabilistic interpretation of feedforward classification network outputs, with relationships to statistical pattern recognition. InNeuro- computing: Algorithms, architectures and applications. Springer, 227–236

    work page 1990

  3. [3]

    Tianqi Chen and Carlos Guestrin. 2016. Xgboost: A scalable tree boosting system. InKDD. 785–794

    work page 2016

  4. [4]

    Yang Chen, Yueheng Jiang, Zhaozhao Ma, Yuchen Cao, Jacky Keung, Kun Kuang, Leilei Gan, Yiquan Wu, and Fei Wu. 2025. MM-DREX: Multimodal- Driven Dynamic Routing of LLM Experts for Financial Trading.arXiv preprint arXiv:2509.05080(2025)

    arXiv 2025

  5. [5]

    Zhiyu Chen, Wenhu Chen, Charese Smiley, Sameena Shah, Iana Borova, Dylan Langdon, Reema Moussa, Matt Beane, Ting-Hao Huang, Bryan R Routledge, et al

  6. [6]

    Zhiyu Chen, Shiyang Li, Charese Smiley, Zhiqiang Ma, Sameena Shah, and William Yang Wang. 2022. Convfinqa: Exploring the chain of numerical reasoning in conversational finance question answering. InEMNLP. 6279–6292

    work page 2022

  7. [7]

    Kyunghyun Cho, Bart Van Merriënboer, Çağlar Gulçehre, Dzmitry Bahdanau, Fethi Bougares, Holger Schwenk, and Yoshua Bengio. 2014. Learning phrase representations using RNN encoder–decoder for statistical machine translation. InEMNLP. 1724–1734

    work page 2014

  8. [8]

    Junyoung Chung, Caglar Gulcehre, KyungHyun Cho, and Yoshua Bengio. 2014. Empirical evaluation of gated recurrent neural networks on sequence modeling. arXiv preprint arXiv:1412.3555(2014)

    arXiv 2014

Show all 65 references
  1. [9]

    Victor DeMiguel, Lorenzo Garlappi, and Raman Uppal. 2009. Optimal versus naive diversification: How inefficient is the 1/N portfolio strategy?The review of Financial studies22, 5 (2009), 1915–1953

    2009

  2. [10]

    Qianggang Ding, Haochen Shi, Jiadong Guo, and Bang Liu. 2024. Tradexpert: Rev- olutionizing trading with mixture of expert llms.arXiv preprint arXiv:2411.00782 (2024)

    2024 arXiv

  3. [11]

    Yujie Ding, Shuai Jia, Tianyi Ma, Bingcheng Mao, Xiuze Zhou, Liuliu Li, and Dongming Han. 2023. Integrating stock features and global information via large language models for enhanced stock return prediction.arXiv preprint arXiv:2310.05627(2023)

    2023 arXiv

  4. [12]

    Zihan Dong, Xinyu Fan, and Zhiyuan Peng. 2024. Fnspid: A comprehensive financial news dataset in time series. InKDD. 4918–4927

    2024

  5. [13]

    Carl Eckart and Gale Young. 1936. The approximation of one matrix by another of lower rank.Psychometrika1, 3 (1936), 211–218

    1936

  6. [14]

    William Fedus, Barret Zoph, and Noam Shazeer. 2022. Switch transformers: Scaling to trillion parameter models with simple and efficient sparsity.Journal of Machine Learning Research23, 120 (2022), 1–39

    2022

  7. [15]

    Fuli Feng, Xiangnan He, Xiang Wang, Cheng Luo, Yiqun Liu, and Tat-Seng Chua

  8. [16]

    Mark B Garman and Michael J Klass. 1980. On the estimation of security price volatilities from historical data.Journal of business(1980), 67–78

    1980

  9. [17]

    Tian Guo and Emmanuel Hauptmann. 2024. Fine-Tuning Large Language Models for Stock Return Prediction Using Newsflow. InEMNLP Industry Track. 1028– 1045

    2024

  10. [18]

    Sepp Hochreiter and Jürgen Schmidhuber. 1997. Long short-term memory.Neural computation(1997)

    1997

  11. [19]

    Kurt Hornik, Maxwell Stinchcombe, and Halbert White. 1989. Multilayer feed- forward networks are universal approximators.Neural networks2, 5 (1989), 359–366

    1989

  12. [20]

    Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Liang Wang, Weizhu Chen, et al. 2022. Lora: Low-rank adaptation of large language models.Iclr1, 2 (2022), 3

    2022

  13. [21]

    Albert Q Jiang, Alexandre Sablayrolles, Antoine Roux, Arthur Mensch, Blanche Savary, Chris Bamford, Devendra Singh Chaplot, Diego de las Casas, Emma Bou Hanna, Florian Bressand, et al . 2024. Mixtral of experts.arXiv preprint arXiv:2401.04088(2024)

    2024 arXiv

  14. [22]

    Zhengyao Jiang, Dixing Xu, and Jinjun Liang. 2017. A deep reinforcement learning framework for the financial portfolio management problem.arXiv preprint arXiv:1706.10059(2017)

    2017 arXiv

  15. [23]

    Guolin Ke, Qi Meng, Thomas Finley, Taifeng Wang, Wei Chen, Weidong Ma, Qiwei Ye, and Tie-Yan Liu. 2017. Lightgbm: A highly efficient gradient boosting decision tree.NeurIPS30 (2017)

    2017

  16. [24]

    Oliver Ledoit and Michael Wolf. 2008. Robust performance hypothesis testing with the Sharpe ratio.Journal of Empirical Finance15, 5 (2008), 850–859

    2008

  17. [25]

    Dmitry Lepikhin, HyoukJoong Lee, Yuanzhong Xu, Dehao Chen, Orhan Firat, Yanping Huang, Maxim Krikun, Noam Shazeer, and Zhifeng Chen. 2021. GShard: Scaling Giant Models with Conditional Computation and Automatic Sharding. InICLR

    2021

  18. [26]

    Tong Li, Zhaoyang Liu, Yanyan Shen, Xue Wang, Haokun Chen, and Sen Huang

  19. [27]

    Hengxu Lin, Dong Zhou, Weiqing Liu, and Jiang Bian. 2021. Learning multiple stock trading patterns with temporal routing adaptor and optimal transport. In Proceedings of the 27th ACM SIGKDD conference on knowledge discovery & data mining. 1017–1026

    2021

  20. [28]

    Aixin Liu, Bei Feng, Bin Wang, Bingxuan Wang, Bo Liu, Chenggang Zhao, Chengqi Dengr, Chong Ruan, Damai Dai, Daya Guo, et al . 2024. Deepseek- v2: A strong, economical, and efficient mixture-of-experts language model.arXiv preprint arXiv:2405.04434(2024)

    2024 arXiv

  21. [29]

    Kuan-Ming Liu and Ming-Chih Lo. 2025. LLM-Based Routing in Mixture of Experts: A Novel Framework for Trading.arXiv preprint arXiv:2501.09636(2025)

    2025 arXiv

  22. [30]

    Xiao-Yang Liu, Hongyang Yang, Jiechao Gao, and Christina Dan Wang. 2021. FinRL: Deep reinforcement learning framework to automate trading in quantita- tive finance. InProceedings of the second ACM international conference on AI in finance. 1–9

    2021

  23. [31]

    Andrew W Lo. 2004. The adaptive markets hypothesis: Market efficiency from an evolutionary perspective.Journal of Portfolio Management, Forthcoming(2004)

    2004

  24. [32]

    Malik Magdon-Ismail, Amir F Atiya, Amrit Pratap, and Yaser S Abu-Mostafa

  25. [33]

    Harry Markowitz. 1952. Portfolio Selection.The Journal of Finance7, 1 (1952), 77–91

    1952

  26. [34]

    Paul Michel, Omer Levy, and Graham Neubig. 2019. Are sixteen heads really better than one?NeurIPS32 (2019)

    2019

  27. [35]

    Pavlo Molchanov, Arun Mallya, Stephen Tyree, Iuri Frosio, and Jan Kautz. 2019. Importance estimation for neural network pruning. InCVPR. 11264–11272

    2019

  28. [36]

    Niklas Muennighoff, Luca Soldaini, Dirk Groeneveld, Kyle Lo, Jacob Morrison, Sewon Min, Weijia Shi, Pete Walsh, Oyvind Tafjord, Nathan Lambert, et al. 2025. Olmoe: Open mixture-of-experts language models. InICLR, Vol. 2025. 62061– 62121

    2025

  29. [37]

    Rajdeep Mukherjee, Abhinav Bohra, Akash Banerjee, Soumya Sharma, Manjunath Hegde, Afreen Shaikh, Shivani Shrivastava, Koustuv Dasgupta, Niloy Ganguly, Saptarshi Ghosh, et al. 2022. Ectsum: A new benchmark dataset for bullet point summarization of long earnings call transcripts...

    2022

  30. [38]

    Karl Pearson. 1895. VII. Note on regression and inheritance in the case of two parents.proceedings of the royal society of London58, 347-352 (1895), 240–242

  31. [39]

    M Hashem Pesaran and Allan Timmermann. 1992. A simple nonparametric test of predictive performance.Journal of Business & Economic Statistics10, 4 (1992), 461–465

    1992

  32. [40]

    Dimitris N Politis and Joseph P Romano. 1994. The stationary bootstrap.Journal of the American Statistical association89, 428 (1994), 1303–1313

    1994

  33. [41]

    William F Sharpe. 1966. Mutual fund performance.The Journal of business39, 1 (1966), 119–138

    1966

  34. [42]

    Noam Shazeer, Azalia Mirhoseini, Krzysztof Maziarz, Andy Davis, Quoc Le, Geoffrey Hinton, and Jeff Dean. 2017. Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layer. InICLR

    2017

  35. [43]

    Yu Shi, Zongliang Fu, Shuo Chen, Bohan Zhao, Wei Xu, Changshui Zhang, and Jian Li. 2026. Kronos: A foundation model for the language of financial markets. InAAAI, Vol. 40. 25366–25373

    2026

  36. [44]

    Yixuan Tang, Yi Yang, Allen H Huang, Andy Tam, and Justin Z Tang. 2023. FinEntity: Entity-level Sentiment Classification for Financial Texts. InEMNLP. 15465–15471

    2023

  37. [45]

    Qwen Team. 2026. Qwen3. 6-35B-A3B: Agentic coding power, now open to all

    2026

  38. [46]

    Chunlin Tian, Zhan Shi, Zhijiang Guo, Li Li, and Chengzhong Xu. 2024. Hydralora: An asymmetric lora architecture for efficient fine-tuning.NeurIPS37 (2024), 9565– 9584

    2024

  39. [47]

    Jingyuan Wang, Yang Zhang, Ke Tang, Junjie Wu, and Zhang Xiong. 2019. Al- phastock: A buying-winners-and-selling-losers investment strategy using inter- pretable deep reinforcement attention networks. InKDD. 1900–1908

    2019

  40. [48]

    Meiyun Wang, Kiyoshi Izumi, and Hiroki Sakaji. 2024. LLMFactor: Extracting Profitable Factors through Prompts for Explainable Stock Movement Prediction. InACL Findings. 3120–3131

    2024

  41. [49]

    Rundong Wang, Hongxin Wei, Bo An, Zhouyan Feng, and Jun Yao. 2021. Com- mission fee is not enough: A hierarchical reinforced framework for portfolio management. InAAAI, Vol. 35. 626–633

    2021

  42. [50]

    1978.New concepts in technical trading systems

    J Welles Wilder. 1978.New concepts in technical trading systems. Greensboro, NC

    1978

  43. [51]

    Shijie Wu, Ozan Irsoy, Steven Lu, Vadim Dabravolski, Mark Dredze, Sebas- tian Gehrmann, Prabhanjan Kambadur, David Rosenberg, and Gideon Mann

  44. [52]

    Yijia Xiao, Edward Sun, Di Luo, and Wei Wang. 2024. TradingAgents: Multi- Agents LLM Financial Trading Framework.arXiv preprint arXiv:2412.20138(2024)

    2024 arXiv

  45. [53]

    Smith, Xiao-Yang Liu, Jimin Huang, Sophia Ananiadou, and Qianqian Xie

    Guojun Xiong, Zhiyang Deng, Keyi Wang, Yupeng Cao, Haohang Li, Yangyang Yu, Xueqing Peng, Mingquan Lin, Kaleb E. Smith, Xiao-Yang Liu, Jimin Huang, Sophia Ananiadou, and Qianqian Xie. 2025. FLAG-TRADER: Fusion LLM-Agent with Gradient-based Reinforcement Learning for Financial ...

    2025

  46. [54]

    Anyi Xu, Bangcai Lin, Bing Xue, Bingxuan Wang, Bingzheng Xu, Bochao Wu, Bowei Zhang, Chaofan Lin, Chen Dong, Chenchen Ling, et al. 2026. Deepseek- v4: Towards highly efficient million-token context intelligence.arXiv preprint arXiv:2606.19348(2026)

    2026

  47. [55]

    Wentao Xu, Weiqing Liu, Lewen Wang, Yingce Xia, Jiang Bian, Jian Yin, and Tie-Yan Liu. 2021. Hist: A graph-based framework for stock trend forecasting via mining concept-oriented shared information.arXiv preprint arXiv:2110.13716 (2021)

    2021 arXiv

  48. [56]

    Hongyang Yang, Xiao-Yang Liu, and Christina Dan Wang. 2023. Fingpt: Open- source financial large language models.arXiv preprint arXiv:2306.06031(2023)

    2023

  49. [57]

    Linyi Yang, Ruihai Dong, Tin Lok James Ng, and Yang Xu. 2019. Leveraging BERT to Improve the FEARS Index for Stock Forecasting. InProceedings of the First Workshop on Financial Technology and Natural Language Processing. Macao, China, 54–60

    2019

  50. [58]

    Yunan Ye, Hengzhi Pei, Boxin Wang, Pin-Yu Chen, Yada Zhu, Ju Xiao, and Bo Li. 2020. Reinforcement-learning based portfolio management with augmented asset movement prediction states. InAAAI, Vol. 34. 1112–1119

    2020

  51. [59]

    Wentao Zhang, Lingxuan Zhao, Haochong Xia, Shuo Sun, Jiaze Sun, Molei Qin, Xinyi Li, Yuqing Zhao, Yilei Zhao, Xinyu Cai, Longtao Zheng, Xinrun Wang, and Bo An. 2024. A Multimodal Foundation Agent for Financial Trading: Tool- Augmented, Diversified, and Generalist. InProceeding...

    2024

  52. [60]

    Zhuohang Zhu, Haodong Chen, Qiang Qu, and Vera Chung. 2025. FinCast: A Foundation Model for Financial Time-Series Forecasting. InProceedings of the 34th ACM International Conference on Information and Knowledge Management. 4539–4549. Conference’17, July 2017, Washington, DC, U...

    2025

  53. [2004]

    On the maximum drawdown of a Brownian motion.Journal of applied probability41, 1 (2004), 147–161

    2004

  54. [2019]

    doi:10.1145/3309547

    Temporal Relational Ranking for Stock Prediction.TOIS37, 2 (2019), 1–30. doi:10.1145/3309547

    2019 doi

  55. [2021]

    Finqa: A dataset of numerical reasoning over financial data. InEMNLP. 3697–3711

  56. [2023]

    Bloomberggpt: A large language model for finance.arXiv preprint arXiv:2303.17564(2023)

    2023 arXiv

  57. [2024]

    In AAAI, Vol

    Master: Market-guided stock transformer for stock price forecasting. In AAAI, Vol. 38. 162–170

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.