Pretrained LLM Adapted with LoRA as a Decision Transformer for Offline RL in Quantitative Trading

Suyeol Yun

REVIEW 4 major objections 5 minor 14 references

Pretrained LLM Adapted with LoRA as a Decision Transformer for Offline RL in Quantitative Trading

T0 review · 4 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read A Decision Transformer initialized with pretrained GPT-2 weights and fine-tuned with LoRA learns trading policies from expert trajectories and consistently beats the same model with random initialization, while matching established…

desk verdict Plausible LaMo-to-finance port, but the never-specified return-to-go deployment protocol could be look-ahead leakage and sink the central claim. read the letter →

arxiv 2411.17900 v1 pith:X4SSEHLL submitted 2024-11-26 q-fin.CP

Suyeol Yun This is my paper · ORCID

classification q-fin.CP

keywords quantitativetradingofflinereinforcementlearningDecisionTransformerLow-RankAdaptation(LoRA)pretrainedlanguagemodelsGPT-2transferDJIA

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

The paper argues that a Decision Transformer—a Transformer trained to predict actions from sequences of returns, states, and past actions—can learn effective trading policies for the 29 stocks of the Dow Jones Industrial Average from a single expert trajectory, provided it starts from pretrained GPT-2 language-model weights and is fine-tuned with Low-Rank Adaptation (LoRA). Across five expert agents (A2C, PPO, SAC, TD3, DDPG), the pretrained-initialized model matches or beats established offline RL baselines (CQL, IQL, BC) on cumulative return, Sharpe ratio, and maximum drawdown, and it consistently outperforms the same architecture with randomly initialized GPT-2 weights. The central empirical claim is that language-model pretraining transfers to financial decision-making, and that LoRA makes this transfer parameter-efficient, with roughly 0.7% of the model's parameters trainable.

What carries the argument

The central mechanism is the Decision Transformer, which reframes reinforcement learning as conditional sequence modeling: each timestep contributes three tokens—return-to-go, state, and action—and the model predicts the next action by minimizing mean squared error over a sliding context window. The paper adapts this architecture to GPT-2 by aligning the input format with GPT-2 token structure, replacing linear embeddings with residual MLPs, freezing the pretrained Transformer weights, and injecting low-rank adapters of rank $r=16$ into the attention layers. The language-pretrained trunk combined with a small set of trainable LoRA parameters (about 900,000, under 1% of GPT-2 small) is what carries the transfer claim: the pretrained attention layers are reused nearly unchanged, so any performance gain over random initialization is attributed to the pretrained representations.

What would settle it

Train the same models on multiple trajectories per expert (for example, different random seeds or stochastic rollouts) and re-evaluate on several non-overlapping test periods; if the pretrained-initialized model no longer consistently beats random initialization, the reported advantage is an artifact of the single-trajectory setup.

Watch

Extended reading notes

Core claim

The paper's central claim is that a Decision Transformer initialized with pretrained GPT-2 weights and fine-tuned using Low-Rank Adaptation (LoRA) learns trading policies from expert trajectories more effectively than the same model with random initialization, and competitively with established offline RL methods. The evidence comes from a simulated DJIA trading environment where five expert RL policies each supply one deterministic trajectory; on the unseen 2020-2021 test window, the pretrained DT-LoRA variant achieves higher cumulative returns than its randomly initialized counterpart in all five cases (for example, 47.98% versus 42.88% under the DDPG expert) and ranks among the top performers on Sharpe ratio and maximum drawdown across baselines. The paper interprets this as evidence that pretrained language representations provide a transferable prior for capturing temporal dependencies in financial time series, while LoRA keeps adaptation computationally feasible.

Load-bearing premise

The load-bearing premise is that a single deterministic expert trajectory is enough to train and compare offline RL methods; if that one path does not cover the state-action distribution well, the test-period rankings, including the pretrained-versus-random gap, may reflect memorization of one policy path rather than general skill.

Editorial extensions

If this is right

If the central claim is correct, pretrained language-model weights provide a reusable initialization for offline RL trading agents, so new trading tasks could be bootstrapped from a language-model prior rather than trained from scratch.
The parameter efficiency of LoRA (under 1% of parameters trainable) suggests that large pretrained models can be adapted to financial domains with limited computational budgets.
The consistent advantage over random initialization indicates that the benefit comes specifically from the pretrained representations, not merely from the Decision Transformer architecture or the LoRA training procedure.
The model's competitive performance against CQL, IQL, and BC, despite training on a single deterministic trajectory, implies that sequence-modeling approaches can extract usable policies from very limited offline data.

Reading between the lines

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

Because the paper trains on only one deterministic trajectory per expert and evaluates on a single post-COVID test window, the observed pretraining advantage could be specific to that market regime; a natural extension is to test across multiple trajectory samples and multiple non-overlapping test periods.
The paper does not isolate what the pretrained weights contribute; ablating which layers are frozen versus adapted, or comparing LoRA against full fine-tuning, would test whether the gain comes from the pretrained trunk or from the embedding architecture.
The same recipe may transfer to other sequential decision problems such as portfolio rebalancing or trade execution, but the paper does not test this.
The small trainable-parameter count raises the possibility that the pretrained initialization acts mainly as a regularizer; comparing against a randomly initialized model with stronger weight-decay or dropout could clarify the mechanism.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

The one thing to know: this paper ports LaMo's pretrained-LM-plus-LoRA Decision Transformer to offline stock trading, and the random-init control is the right experiment. But the paper never says how return-to-go (RTG) is set at test time, and if the model sees ground-truth future returns, every number in Table 2 collapses. That is not a corner-case worry; it is load-bearing.

What is genuinely here: a clean application of an existing recipe to DJIA trading, with residual MLP embeddings and parameter-matched baselines. The pretrained-versus-random comparison is a sensible way to isolate the value of GPT-2 initialization. The prose is clear, the code is public, and the paper honestly credits LaMo as the framework it adopts. That is all real.

Now the soft spots, in proportion. The RTG issue is the big one. In a Decision Transformer, the policy is conditioned on RTG; at deployment you need an initial target and an update rule. Section 3.2 defines RTG as the sum of future rewards; the evaluation sections (4.2, 4.4, A.11) describe only "same environment settings" with unseen data. No inference-time RTG protocol is given. If the code feeds the expert's realized test-period returns or the ground-truth future sequence, the model has look-ahead information and the pretrained-vs-random gap in Figure 3 could reflect leakage rather than transferable prior knowledge. The repository can settle this.

The single deterministic trajectory per expert (Appendix A.10) and the single July 2020-October 2021 test window make the generalization claims thin. The paper also overstates: "significantly outperforms" is used without significance tests, and several comparisons have overlapping error bars (e.g., SAC returns: 39.59±3.18 vs. 37.07±0.55). The prose claims "consistently outperforms" when the evidence is suggestive, not conclusive.

The paper is not muddled. It acknowledges the LaMo lineage, uses a fair control, and releases code. It is a credible workshop-level application, not a new mechanism. Who gets value: fin-RL practitioners wanting a starting point for pretrained-LM priors. It deserves peer review because the RTG ambiguity is fixable and the empirical question is worth answering rigorously. My recommendation: send to review, but require the authors to specify the deployment-time RTG rule, add a non-leakage baseline (e.g., a fixed target), and ideally more trajectories and test windows. Without that, the headline result is unverifiable.

Referee Report

4 major / 5 minor

Summary. The paper proposes a Decision Transformer (DT) initialized with pretrained GPT-2 weights and fine-tuned with Low-Rank Adaptation (LoRA) for offline reinforcement learning in quantitative trading. The method is trained on expert trajectories from five RL agents (A2C, PPO, SAC, TD3, DDPG) in a FinRL DJIA trading environment and evaluated on a single test window (July 2020 to October 2021). The authors compare against CQL, IQL, BC, and a randomly initialized DT with LoRA, reporting cumulative return, maximum drawdown, and Sharpe ratio. The central claim is that the GPT-2-initialized DT performs competitively with established offline RL methods and consistently outperforms its randomly initialized counterpart.

Significance. If validated, the result would be a meaningful empirical contribution: it would suggest that pretrained language model weights provide a transferable prior for learning trading policies from expert trajectories, and the random-initialization control is a sensible experimental design for isolating the effect of pretrained weights. The public replication code is a concrete strength. However, the significance is contingent on the deployment-time return-to-go protocol being causal, on the single-trajectory dataset supporting the offline RL claim, and on the statistical robustness of the reported comparisons. As written, those conditions are not established, so the empirical contribution is currently unverified.

major comments (4)

[Section 3.2, Eq. (4); Section 4.2 and Appendix A.11] The deployment-time return-to-go (RTG) protocol is never specified. The model is trained with RTG defined as the sum of future rewards (Eq. 4) and the action prediction is conditioned on this quantity. At test time, a Decision Transformer requires an initial RTG target and an update rule based on realized rewards; the manuscript only states that evaluation uses 'the same environment settings' with unseen data (A.11). If the test-time RTG is initialized with the expert's realized test-period cumulative return, or if ground-truth future returns for the test window are fed as inputs, then the model has look-ahead information about the evaluation period. This would make every number in Table 2 and Figure 3, including the pretrained-versus-random gap, reflect leakage rather than learned trading skill. Please specify the exact RTG initialization and update protocol used at deployment, and if the code uses realized test-period returns, rerun the evaluation with a causal RTG schedule.
[Appendix A.10] Only one deterministic trajectory per expert is sampled for the offline dataset. This means the training data is a single path through the state-action space, with no coverage of alternative states or actions. The models are therefore trained to reproduce one policy rollout, which is closer to memorizing a single trajectory than to learning an offline RL policy from a diverse dataset. The pretrained-versus-random comparison may then reflect which initialization memorizes that one path better, not a transferable prior for trading. Please collect multiple stochastic (or otherwise diverse) trajectories per expert and report results, or explicitly reframe the claims as behavior cloning from a single expert rollout.
[Table 2 and Figure 3] No significance tests, confidence intervals, or per-seed breakdowns are reported. Appendix A.7 lists five random seeds but does not state how they are used or how many runs underlie each mean and standard deviation. In the PPO row, the pretrained model's cumulative return is 34.99 ± 8.03 versus 28.76 ± 0.27 for the random initialization, so the reported variability is large enough to undermine the claim of a 'consistent performance boost.' Report the number of runs, per-seed results, and pairwise significance tests or bootstrap confidence intervals for all three metrics across all five expert agents.
[Section 4.2] The evaluation uses a single test window, July 2020 to October 2021, which is one post-COVID market regime. The abstract and Section 4.5 claim generalization to unseen market conditions, but one test period cannot establish robustness across market regimes. Add at least one additional out-of-sample period (for example 2022-2023) or temper the generalization claims to this specific test window.

minor comments (5)

[Section 4.3] The text contains a literal formatting artifact, 'vbnet Copy code', interrupting the experimental design subsection. This should be removed.
[References] References [2] and [3] cite the same Decision Transformer paper (arXiv:2106.01345 and the NeurIPS version), and references [6] and [7] cite the same LoRA paper. These duplicates should be consolidated.
[Figure 3] The bar charts report only point estimates, while Table 2 provides standard deviations. Add error bars to Figure 3 so the visual comparison reflects the underlying variability.
[Appendix A.7] Five random seeds are listed, but the manuscript does not explain whether each experiment was run once per seed or whether these seeds correspond to different training runs for each expert agent and method. Clarify the mapping of seeds to runs.
[Section 4.5.2] The phrase 'reflecting a significant performance gain' for the DDPG case (47.98% versus 42.88%) uses 'significant' in a statistical sense that is not supported by any test; rephrase to avoid implying formal significance.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the central pretrained-versus-random comparison is an independent controlled experiment, not a fitted quantity renamed as a prediction.

full rationale

The paper's core claims are empirical: a GPT-2-initialized Decision Transformer with LoRA is compared on a held-out test window against established offline RL baselines and against a randomly initialized control. The comparison is not derived from the paper's inputs by construction: the DT models are trained by action-MSE cloning of expert trajectories over 2009-2020 and evaluated on 2020-2021 data, so the test-period cumulative returns, Sharpe ratios, and MDDs are genuine out-of-sample predictions. The LoRA/GPT-2 recipe is adopted from LaMo [12] by citation, but the paper does not invoke LaMo to justify the success claim; the success claim rests on Table 2 and Figure 3. There are no self-citations by the single author, no imported uniqueness theorem, and no parameter fitted to test data. The only noted concern, the unspecified deployment-time return-to-go protocol (Section 3.2 versus Appendix A.11), would be a look-ahead validity issue if realized, but the paper provides no text showing that RTG is set to ground-truth future returns at test time, so it cannot be counted as a circular reduction under the evidence rule. Overall, the derivation chain is self-contained and externally falsifiable.

Assumptions & free parameters 7 free parameters · 6 assumptions · 0 invented entities

The central claim rests on several unverified domain assumptions: the FinRL environment faithfully models trading, a single expert trajectory is a sufficient dataset, the 2020-2021 test window is representative, and GPT-2 pretrained weights transfer to numerical financial tokens. Hyperparameters are hand-set without a reported validation sweep. No new entities are introduced.

free parameters (7)

Context length K = 20
Appendix A.1; chosen without a reported sweep.
LoRA rank r = 16
Appendix A.4; chosen without a reported sweep.
Learning rate = 1e-3
Appendix A.2; fixed across all models.
Batch size = 64
Appendix A.2; fixed across all models.
Training iterations = 1000
Appendix A.2; fixed across all models.
Weight decay = 1e-5
Appendix A.2; fixed across all models.
Embedding hidden size = 768
Appendix A.1; set to GPT-2 hidden size.

assumptions (6)

domain assumption The FinRL DJIA environment is a valid MDP model of trading.
Sections 2.1 and 4.1; the whole experiment assumes states, actions, and rewards from FinRL are representative of real market dynamics.
ad hoc to paper A single deterministic expert trajectory is a sufficient offline dataset.
Appendix A.10 explicitly samples one trajectory per agent; this is load-bearing for the comparison and is not justified.
domain assumption GPT-2 pretrained representations transfer to numerical financial token sequences.
Section 3 and Table 1; the method assumes language pretraining helps financial sequence modeling, tested only indirectly by the random-init control.
domain assumption The July 2020 to October 2021 window is representative for generalization.
Section 4.2; all conclusions are drawn from this single test period.
standard math MSE action prediction is a valid objective for offline policy extraction.
Equation (21), the standard Decision Transformer objective.
domain assumption LoRA low-rank updates are a sufficient adapter for the pretrained weights.
Sections 2.3 and A.4; no comparison to full fine-tuning is provided.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Pretrained LLM Adapted with LoRA as a Decision Transformer for Offline RL in Quantitative Trading." pith.science (2026). https://pith.science/paper/X4SSEHLL

@misc{pith2026241117900,
  author       = {Pith},
  title        = {Pith review of: Pretrained LLM Adapted with LoRA as a Decision Transformer for Offline RL in Quantitative Trading},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/X4SSEHLL}},
  note         = {Machine review of arXiv:2411.17900}
}

read the original abstract

Developing effective quantitative trading strategies using reinforcement learning (RL) is challenging due to the high risks associated with online interaction with live financial markets. Consequently, offline RL, which leverages historical market data without additional exploration, becomes essential. However, existing offline RL methods often struggle to capture the complex temporal dependencies inherent in financial time series and may overfit to historical patterns. To address these challenges, we introduce a Decision Transformer (DT) initialized with pre-trained GPT-2 weights and fine-tuned using Low-Rank Adaptation (LoRA). This architecture leverages the generalization capabilities of pre-trained language models and the efficiency of LoRA to learn effective trading policies from expert trajectories solely from historical data. Our model performs competitively with established offline RL algorithms, including Conservative Q-Learning (CQL), Implicit Q-Learning (IQL), and Behavior Cloning (BC), as well as a baseline Decision Transformer with randomly initialized GPT-2 weights and LoRA. Empirical results demonstrate that our approach effectively learns from expert trajectories and secures superior rewards in certain trading scenarios, highlighting the effectiveness of integrating pre-trained language models and parameter-efficient fine-tuning in offline RL for quantitative trading. Replication code for our experiments is publicly available at https://github.com/syyunn/finrl-dt

Figures

Figures reproduced from arXiv: 2411.17900 by the authors.

**Figure 2.** Comparison of cumulative returns over time for different offline RL methods trained on A2C expert trajectories. [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

**Figure 3.** Performance Comparison between DT-LoRA-GPT2 with Pre-trained Weights and Random Initialization across [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗

Figures from the paper (5 more)

**Figure 3.** Figure 3: • Cumulative Return: The model achieves higher cumulative returns in every case. For example, under DDPG, DT-LoRAGPT2 attains a return of 47.98% versus 42.88% for the random initialization, reflecting a significant performance gain. • Sharpe Ratio: DT-LoRA-GPT2 genera…

**Figure 4.** Figure 4: Performance Comparison for PPO Expert Agent [PITH_FULL_IMAGE:figures/full_fig_p011_4.png]

**Figure 5.** Figure 5: Performance Comparison for TD3 Expert Agent [PITH_FULL_IMAGE:figures/full_fig_p011_5.png]

**Figure 6.** Figure 6: Performance Comparison for SAC Expert Agent [PITH_FULL_IMAGE:figures/full_fig_p012_6.png]

**Figure 7.** Figure 7: Performance Comparison for DDPG Expert Agent [PITH_FULL_IMAGE:figures/full_fig_p012_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

14 extracted references · 6 canonical work pages

[1]

Wang, and Longbing Cao

Bo An, Shuo Sun, R. Wang, and Longbing Cao. 2022. Deep Reinforcement Learning for Quantitative Trading: Challenges and Opportunities.IEEE Intelligent Systems 37 (2022), 23–26. https://api.semanticscholar.org/CorpusID:248922126

work page 2022
[2]

Lili Chen, Kevin Lu, Aravind Rajeswaran, Kimin Lee, Aditya Grover, Michael Laskin, Pieter Abbeel, Aravind Srinivas, and Igor Mordatch. 2021. Decision Transformer: Reinforcement Learning via Sequence Modeling. arXiv:2106.01345 [cs.LG] https://arxiv.org/abs/2106.01345 Pretrained LLM Adapted with LoRA as a DT for Offline RL in Trading ICAIF ’24 Workshop LLM ...

arXiv 2021
[3]

Lili Chen, Kevin Lu, Aravind Srinivas, Tsung-Yi Lin, Pieter Abbeel, and Igor Mordatch. 2021. Decision Transformer: Reinforcement Learning via Sequence Modeling. In Proceedings of the 35th Conference on Neural Information Processing Systems (NeurIPS)

work page 2021
[4]

Gabriel Dulac-Arnold, Daniel Jaymin Mankowitz, and Todd Hester. 2019. Chal- lenges of Real-World Reinforcement Learning. ArXiv abs/1904.12901 (2019). https://api.semanticscholar.org/CorpusID:140269588

arXiv 2019
[5]

Thomas G. Fischer. 2018. Reinforcement learning in financial markets - a survey. https://api.semanticscholar.org/CorpusID:169730324

work page 2018
[7]

Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, and Weizhu Chen

Edward J. Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, and Weizhu Chen. 2021. LoRA: Low-Rank Adaptation of Large Language Models. arXiv preprint arXiv:2106.09685 (2021)

arXiv 2021
[8]

Sergey Levine, Aviral Kumar, George Tucker, and Justin Fu. 2020. Offline Rein- forcement Learning: Tutorial, Review, and Perspectives on Open Problems.CoRR abs/2005.01643 (2020). arXiv:2005.01643 https://arxiv.org/abs/2005.01643

arXiv 2020
[9]

Xiao-Yang Liu, Ziyi Xia, Jingyang Rui, Jiechao Gao, Hongyang Yang, Ming Zhu, Christina Wang, Zhaoran Wang, and Jian Guo. 2022. FinRL-Meta: Market En- vironments and Benchmarks for Data-Driven Financial Reinforcement Learn- ing. In Advances in Neural Information Processing Systems , S. Koyejo, S. Mo- hamed, A. Agarwal, D. Belgrave, K. Cho, and A. Oh (Eds.)...

work page 2022

Show all 14 references

[10]

Yang Liu, Qi Liu, Hongke Zhao, Zhen Pan, and Chuanren Liu. 2020. Adaptive Quantitative Trading: An Imitative Deep Reinforcement Learning Approach. Proceedings of the AAAI Conference on Artificial Intelligence 34, 02 (Apr. 2020), 2128–2135. https://doi.org/10.1609/aaai.v34i02.5587

2020 doi
[11]

Alec Radford, Jeff Wu, Rewon Child, David Luan, Dario Amodei, and Ilya Sutskever. 2019. Language Models are Unsupervised Multitask Learners. https: //api.semanticscholar.org/CorpusID:160025533

2019
[12]

Du, and Huazhe Xu

Ruizhe Shi, Yuyao Liu, Yanjie Ze, Simon S. Du, and Huazhe Xu. 2023. Unleashing the Power of Pre-trained Language Models for Offline Reinforcement Learning. arXiv:2310.20587 [cs.LG] https://arxiv.org/abs/2310.20587

2023 arXiv
[13]

Shuo Sun, Rundong Wang, and Bo An. 2023. Reinforcement Learning for Quanti- tative Trading. ACM Trans. Intell. Syst. Technol. 14, 3, Article 44 (March 2023), 29 pages. https://doi.org/10.1145/3582560

2023 doi
[14]

Gomez, Lukasz Kaiser, and Illia Polosukhin

Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser, and Illia Polosukhin. 2017. Attention Is All You Need. CoRR abs/1706.03762 (2017). arXiv:1706.03762 http://arxiv.org/abs/ 1706.03762

2017 arXiv
[15]

Chuheng Zhang, Yitong Duan, Xiaoyu Chen, Jianyu Chen, Jian Li, and Li Ping Zhao. 2023. Towards Generalizable Reinforcement Learning for Trade Execu- tion. ArXiv abs/2307.11685 (2023). https://api.semanticscholar.org/CorpusID: 260091562 ICAIF ’24 Workshop LLM & GenAI for Financ...

2023 arXiv

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.

[1] [1]

Wang, and Longbing Cao

Bo An, Shuo Sun, R. Wang, and Longbing Cao. 2022. Deep Reinforcement Learning for Quantitative Trading: Challenges and Opportunities.IEEE Intelligent Systems 37 (2022), 23–26. https://api.semanticscholar.org/CorpusID:248922126

work page 2022

[2] [2]

Lili Chen, Kevin Lu, Aravind Rajeswaran, Kimin Lee, Aditya Grover, Michael Laskin, Pieter Abbeel, Aravind Srinivas, and Igor Mordatch. 2021. Decision Transformer: Reinforcement Learning via Sequence Modeling. arXiv:2106.01345 [cs.LG] https://arxiv.org/abs/2106.01345 Pretrained LLM Adapted with LoRA as a DT for Offline RL in Trading ICAIF ’24 Workshop LLM ...

arXiv 2021

[3] [3]

Lili Chen, Kevin Lu, Aravind Srinivas, Tsung-Yi Lin, Pieter Abbeel, and Igor Mordatch. 2021. Decision Transformer: Reinforcement Learning via Sequence Modeling. In Proceedings of the 35th Conference on Neural Information Processing Systems (NeurIPS)

work page 2021

[4] [4]

Gabriel Dulac-Arnold, Daniel Jaymin Mankowitz, and Todd Hester. 2019. Chal- lenges of Real-World Reinforcement Learning. ArXiv abs/1904.12901 (2019). https://api.semanticscholar.org/CorpusID:140269588

arXiv 2019

[5] [5]

Thomas G. Fischer. 2018. Reinforcement learning in financial markets - a survey. https://api.semanticscholar.org/CorpusID:169730324

work page 2018

[6] [7]

Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, and Weizhu Chen

Edward J. Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, and Weizhu Chen. 2021. LoRA: Low-Rank Adaptation of Large Language Models. arXiv preprint arXiv:2106.09685 (2021)

arXiv 2021

[7] [8]

Sergey Levine, Aviral Kumar, George Tucker, and Justin Fu. 2020. Offline Rein- forcement Learning: Tutorial, Review, and Perspectives on Open Problems.CoRR abs/2005.01643 (2020). arXiv:2005.01643 https://arxiv.org/abs/2005.01643

arXiv 2020

[8] [9]

Xiao-Yang Liu, Ziyi Xia, Jingyang Rui, Jiechao Gao, Hongyang Yang, Ming Zhu, Christina Wang, Zhaoran Wang, and Jian Guo. 2022. FinRL-Meta: Market En- vironments and Benchmarks for Data-Driven Financial Reinforcement Learn- ing. In Advances in Neural Information Processing Systems , S. Koyejo, S. Mo- hamed, A. Agarwal, D. Belgrave, K. Cho, and A. Oh (Eds.)...

work page 2022