Pith. sign in

REVIEW 4 major objections 5 minor 29 references

CLaC@FinMMEval 2026 Task 3: Sentiment-Augmented Deep Reinforcement Learning for Active Trading -- An Alpha-Reward Approach

T0 review · 4 major / 5 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read The central claim is that replacing the raw log-return reward with a market-relative alpha reward — the day's portfolio growth minus the day's market growth — aligns training with beating buy-and-hold and, paired with random episode starts,

desk verdict Solid, honest competition write-up with a useful regime-shift result; the alpha reward is mathematically trivial and its overfitting-reduction claim is unsubstantiated, but the paper deserves a referee as working notes. read the letter →

arxiv 2607.16028 v1 pith:2KE6KALF submitted 2026-07-17 cs.LG

Andrei Neagu , Eeham Khan , Leila Kosseim This is my paper · ORCID
classification cs.LG
keywords deepreinforcementlearningalpharewardalgorithmictradingbuy-and-holdsentimentanalysisDDPGmarketregimeshiftexcessreturn
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 tries to establish that a trading agent should optimize excess return over the market rather than absolute return. To this end it introduces an alpha reward, which subtracts the daily log return of the asset from the daily log return of the agent's portfolio. The paper proves three properties of this reward: summed over an episode it equals log-alpha (the log of the ratio of portfolio growth to market growth); under the price-taker assumption it shares its optimal policies with the raw log-return reward; and it acts as a variance-reducing control variate. Empirically, a DDPG agent using this reward achieves a 54.96% cumulative return on TSLA versus 16.45% for buy-and-hold, and a 1.58% return on BTC versus -34.27% for buy-and-hold. The paper also documents a large validation-to-test generalization gap, concluding that validation-based model selection is fragile when the validation and test periods are in different market regimes.

What carries the argument

The central object is the alpha reward, a market-relative reward defined as the difference between the daily log portfolio return and the daily log market return. It does three jobs: (1) it telescopes so that the episode return equals log-alpha, the log of the ratio of final portfolio value to final buy-and-hold value; (2) under the price-taker assumption it leaves the set of optimal policies unchanged relative to the raw log-return reward, since the subtracted market term is policy-independent; and (3) it acts as a control variate, stripping market-driven variance from the reward signal. The other machinery is random episode start dates, which forces the agent to experience diverse market c

What would settle it

Evaluate the same trained policies on the test window with the 0.2% transaction cost applied at each position change (as it is during training) instead of zero cost. If DDPG's TSLA cumulative return drops below buy-and-hold, or its BTC return turns negative, the claim that the alpha-reward DDPG beats buy-and-hold is refuted. Alternatively, train a raw-log-return agent with identical seeds and hyperparameters; theory says the optimal policies coincide, so if the alpha-reward agent's test returns are not statistically distinguishable from the raw-reward agent's across several seeds, the variance

Watch

Extended reading notes

Core claim

The paper's central discovery is the alpha reward: at each day t, r_t = log(V_t/V_{t-1}) - log(c_t/c_{t-1}), where V_t is the portfolio value and c_t is the asset's closing price. With this reward a buy-and-hold strategy receives zero reward every step, so the agent is trained explicitly to outperform the market. The reward telescopes over an episode to the log of the terminal outperformance ratio alpha_T, shares its optimal policies with the raw log-return objective whenever market dynamics do not depend on the policy (the price-taker assumption), and reduces the variance of the gradient estimate because the market component is subtracted out. In the paper's experiments, DDPG trained with t

Load-bearing premise

The headline results rest on a single realized test path evaluated without transaction costs, so if that path is unrepresentative or costs are reintroduced at inference, the ordering of agents and even the sign of the BTC return could change.

Editorial extensions

If this is right

  • If the alpha reward is adopted, training objective and evaluation criterion coincide: an agent is graded on outperforming buy-and-hold, not on absolute profit, so a profitable market does not mask a bad policy.
  • The policy-equivalence result implies that any performance gain from the alpha reward in practice comes from variance reduction and better credit assignment, not from a changed optimum; algorithms that struggle with high-variance rewards should benefit most.
  • DDPG, a continuous-action actor-critic with replay, proved more stable than recurrent on-policy methods (PG, PPO) and value-based DQL, suggesting that off-policy stability matters more than temporal feature processing for this task.
  • Sharpe-ratio-based model selection on a single bull-market validation window does not guarantee test performance in a bear market; the paper's own results imply that walk-forward or regime-diverse validation is needed for reliable selection.
  • The framework should transfer to other long/flat/short trading tasks, since the alpha reward is asset-agnostic and needs only the asset's daily return as a benchmark.

Reading between the lines

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

  • Because the alpha reward shares optimal policies with the raw log-return reward, a direct ablation comparing two otherwise identical agents — one trained with each reward — would isolate whether the empirical gains come from variance reduction, and this comparison should be run across multiple seeds.
  • The paper evaluates without transaction costs at test time; re-running the same checkpoints with the 0.2% cost applied at inference would test whether the reported outperformance survives realistic frictions, which the authors themselves flag as a limitation.
  • The sentiment score's marginal contribution is untested; a natural follow-up is to retrain with sentiment features zeroed out, or replaced by a lexicon baseline, to see whether the results are driven by price-derived indicators rather than news.
  • The regime-shift finding suggests a practical selection rule: choose the model with the best validation Sharpe across multiple non-overlapping windows spanning both bull and bear phases, rather than a single recent window.
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

4 major / 5 minor

Summary. The paper presents a system for CLEF 2026 FinMMEval Task 3, formulating daily long/flat/short trading as a discrete-action MDP. It compares PG, PPO, DQL, and DDPG on a feature set of technical indicators, calendar encodings, and LLaMA 3.2-based news sentiment. The methodological contribution is an 'alpha reward' r_t = log(V_t/V_{t-1}) - log(c_t/c_{t-1}) combined with random episode starts, claimed to reduce overfitting and align training with the evaluation metric of beating buy-and-hold. On the test set, DDPG achieves the best reported performance (TSLA CR 54.96%, SR 1.44; BTC CR 1.58%, SR 0.23), while DQL was selected a priori for the live endpoint based on validation Sharpe ratio. The paper also documents a large validation-to-test generalization gap, attributed to regime shift.

Significance. The alpha-reward telescoping identity (Eq. 4) and the equivalence of optimal policies with raw log-return reward (Property 2) are correctly stated under the price-taker assumption. The paper is transparent about its a-priori endpoint selection and about the train/test gap, and the reproducibility plan is detailed. However, the central claim that the alpha reward 'reduces overfitting' is not established: the variance-reduction argument is not valid for the full discrete action set, and no ablation isolates the reward change from random episode starts. The headline empirical results rest on one cost-free test path without confidence intervals. These issues are load-bearing for the paper's main contribution, though they are fixable with additional experiments.

major comments (4)
  1. [§4.4, Property 3; §1 Contribution 2] The variance-reduction/overfitting-reduction mechanism is not valid for the full action set. For a flat position, r_t^alpha = -m_t, so Var(r^alpha)=Var(m_t) while the raw reward has zero variance. For a short position, to first order r_t^alpha ≈ -2m_t versus r_t^raw ≈ -m_t, so subtracting the market return doubles the market-noise variance. Property 3 holds only if the policy is almost always long. No ablation separates the alpha reward from random episode starts; §7 lists such an ablation as future work. Moreover, §6.1 reports training returns of 114,462% (TSLA) and 256,989% (BTC), indicating severe overfitting persists. The statement that the alpha reward 'reduces overfitting' is therefore unsupported as written.
  2. [§5.2 and Table 4] The headline comparison rests on a single realized test path. No multiple seeds, bootstrap confidence intervals, or seed-variance information are reported. Evaluation is cost-free, as transaction costs are applied only during training, and episodes always begin from the first day of the period. Consequently the ranking DDPG > DQL > buy-and-hold, and even the sign of BTC returns, may change under realistic inference-time costs or a different test path. The paper should report uncertainty estimates (e.g., repeated seeds or resampling) and qualify the 'most robust' characterization of DDPG.
  3. [§5.1 and §5.2] The hyperparameter search runs 180 trials per algorithm-asset pair, but the random seed protocol for final training and evaluation is not described. With a single seed, the differences between algorithms (DQL vs DDPG on TSLA, or PG vs PPO on BTC) may be within optimization noise. Please state the number of seeds used and, ideally, report the mean and spread over seeds, or explicitly treat the reported numbers as single-run observations.
  4. [§6.2] The paper correctly identifies the bull-market validation window and bear-market test window as the source of the generalization gap, and admits walk-forward cross-validation was not implemented due to computational cost. This is an honest limitation, but it means the empirical conclusions are specific to one temporal split and one test path. The claims about 'robust performance' should be narrowed accordingly, or supported by at least a small walk-forward or regime-shift analysis.
minor comments (5)
  1. [§4.1] 'articles published after 16:00 UTC (i.e., after the U.S. equity market close)' is factually incorrect; the U.S. close is approximately 20:00/21:00 UTC. The shift is conservative, not look-ahead, but the text should be corrected to avoid confusion about the alignment rule.
  2. [Table 2] Rows 9 and 10 use EMA10/c_{t-1} and EMA50/c_{t-1} as formulas, while the text says each EMA is normalized by the current close. Please align notation.
  3. [§4.3] The heading 'log-probability extraction' is a misnomer: the method extracts logits and applies a restricted softmax. Consider renaming to 'logit-based probability extraction'.
  4. [§5.2] The sentence 'model selection uses the Sharpe ratio computed on the raw portfolio value series' is ambiguous; clarify that 'raw' means the unadjusted portfolio value, not the raw reward.
  5. [Figures 3 and 4] Captions describe portfolio values but the y-axis appears to be in dollars; please state the assumed initial capital for the plots.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; alpha-reward identities are definitional but not used as independent evidence, and test evaluation is blind to the test set.

full rationale

The paper's derivation chain is self-contained and non-circular. Equation (3) defines the alpha reward as excess log-return; Equation (4) is a telescoping identity that follows immediately from this definition, and Property 2 (equivalent optima) follows directly from the stated price-taker assumption that market returns are policy-independent. Neither step imports the conclusion or fits the target quantity. The empirical claims rest on a held-out CLEF test period; hyperparameters were selected on a validation Sharpe ratio, the live endpoint model (DQL) was chosen a priori and blind to the test period, and DDPG's headline returns are reported as offline backtest results rather than post-hoc fitted predictions. The paper contains no load-bearing self-citations by its authors; citations to the FinMMEval task overviews describe the external benchmark. The variance-reduction Property 3 is not generally valid in the discrete-action setting (e.g., for flat positions it replaces zero raw-reward variance with Var(m_t), and for short positions it amplifies market noise), and the 'reduces overfitting' contribution is not supported by an ablation; however, these are correctness/evidence limitations, not cases where a claimed prediction reduces by construction to its input. Therefore no circularity step meets the evidentiary bar.

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

The central empirical claim rests on standard RL/finance modeling assumptions: price-taking, state sufficiency, validity of LLM sentiment, and informativeness of the validation split. The alpha reward is not a new entity but a reward transformation. The free parameters are mainly hyperparameters and the DDPG discretization threshold, none of which are reported with their selected values.

free parameters (4)
  • Hyperparameters (hidden_dim, num_layers, gamma, lr, entropy_coeff, clip_epsilon, tau) = not reported (selected per algorithm-asset pair by Ray Tune over 180 trials)
    All final results depend on validation-Sharpe-based hyperparameter selection; the best configurations are not listed in the paper.
  • DDPG action discretization thresholds (±0.33) = ±1/3
    Chosen by hand to equal-width partition the continuous action range; affects the mapping from continuous actor output to discrete long/flat/short actions.
  • Training transaction cost δ = 0.002
    Applied during training only; shapes the inertia incentive and is a free modeling choice relative to the cost-free evaluation protocol.
  • Early stopping patience = 10 validation evaluations, each 50 episodes apart
    Affects which checkpoint is selected for testing; chosen without sensitivity analysis.
assumptions (5)
  • domain assumption Price-taker assumption: agent actions do not influence future market states (Section 3, P(s'|s,a)).
    This is the key premise for the alpha-reward equivalence to raw log-return optima. It is standard for small traders but not tested empirically.
  • domain assumption The 15-dimensional state vector (technical indicators, calendar encodings, sentiment, prior position) is a sufficient summary of decision-relevant information.
    No ablation is provided to quantify the marginal contribution of any feature group; Section 7 explicitly calls for such an ablation.
  • domain assumption LLaMA 3.2 1B zero-shot log-probability sentiment scores are a valid news signal.
    No comparison against FinBERT, lexicon baselines, or human-labeled sentiment is provided; Section 7 lists replacing the model as future work.
  • domain assumption The single chronological validation window (2023-2024) is informative for model selection.
    The paper itself shows this fails for BTC (bull validation vs bear test); walk-forward cross-validation was not implemented due to time constraints (Section 6.2).
  • domain assumption External data sources (Yahoo Finance prices/volumes, Brian Ferrell news dataset) are accurate and temporally aligned without look-ahead.
    The temporal alignment rules are described, but the correctness of the external data itself is taken as given.

how reviews work

0 comments
Cite this review

Pith. "Pith review of CLaC@FinMMEval 2026 Task 3: Sentiment-Augmented Deep Reinforcement Learning for Active Trading -- An Alpha-Reward Approach." pith.science (2026). https://pith.science/paper/2KE6KALF

@misc{pith2026260716028,
  author       = {Pith},
  title        = {Pith review of: CLaC@FinMMEval 2026 Task 3: Sentiment-Augmented Deep Reinforcement Learning for Active Trading -- An Alpha-Reward Approach},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2KE6KALF}},
  note         = {Machine review of arXiv:2607.16028}
}
read the original abstract

This paper presents our system for Task 3 of the CLEF 2026 FinMMEval Lab, which requires daily long, flat, or short trading decisions for Bitcoin (BTC) and Tesla (TSLA) using news and historical market data. We formulate the problem as a discrete-action Markov Decision Process and compare four deep reinforcement learning algorithms: Policy Gradient (PG), Proximal Policy Optimization (PPO), Deep Q-Learning (DQL), and Deep Deterministic Policy Gradient (DDPG). The agents use technical indicators, cyclical calendar encodings, and daily news sentiment scores produced by LLaMA 3.2 1B. To reduce overfitting and align training with the objective of outperforming buy-and-hold, we introduce an alpha reward based on excess market return and randomize episode start dates. Hyperparameters are optimized with Ray Tune over 180 trials per algorithm-asset pair, with early stopping and model selection based on validation Sharpe ratio. On the CLEF Task 3 test set, DDPG achieves the strongest overall performance. DQL was selected a priori for the live endpoint because it obtained the highest validation Sharpe ratio, with selection performed without access to the test period. For TSLA, DDPG and DQL achieve cumulative returns of 54.96% and 52.62%, respectively, compared with 16.45% for buy-and-hold. For BTC, DDPG achieves a positive return of 1.58% while buy-and-hold declines by -34.27%. The results also reveal a substantial validation-to-test generalization gap, highlighting the difficulty of transferring policies selected in bull-market conditions to a bear-market regime.

Figures

Figures reproduced from arXiv: 2607.16028 by the authors.

Figure 1
Figure 1. System overview. Left: three parallel processing streams, technical indicators derived from price and volume (purple) and daily news sentiment from a frozen LLaMA-3.2-1B Instruct model (blue), together with cyclical calendar encodings, compose a 15-dimensional state vector 𝑠𝑡. The state is fed as input by a custom trading environment whose 𝛼-reward incentivises outperforming the buy-and-hold baseline, and by one of … view at source ↗
Figure 2
Figure 2. Asset prices (in log scale) across time and data splits. (a) TSLA prices (in log scale) across time and data splits. (b) BTC prices (in log scale) across time and data splits. As shown in [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. DDPG vs B&H portfolio values on the TSLA and BTC test sets. (a) DDPG vs B&H portfolio values on the TSLA test set. (b) DDPG vs B&H portfolio values on the BTC test set. On TSLA (Figure 3a), DDPG leads buy-and-hold throughout most of the test period. Both series rise together from August to December 2025, peaking over $160, 000 for buy-and-hold. After that peak, buy-and-hold retraces sharply as TSLA gives back gains,… view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: DQL vs B&H portfolio values on the TSLA and BTC test sets. (a) DQL vs B&H portfolio values on the TSLA test set. (b) DQL vs B&H portfolio values on the BTC test set. On TSLA (Figure 4a), DQL takes an initially incorrect short position while TSLA rises, falling to rough…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

29 extracted references · 1 canonical work pages

  1. [1]

    V. Mnih, K. Kavukcuoglu, D. Silver, A. A. Rusu, J. Veness, M. G. Bellemare, A. Graves, M. Riedmiller, A. K. Fidjeland, G. Ostrovski, S. Petersen, C. Beattie, A. Sadik, I. Antonoglou, H. King, D. Kumaran, D. Wierstra, S. Legg, D. Hassabis, Human-level control through deep reinforcement learning, Nature 518 (2015) 484–489

    2015

  2. [2]

    Z. Xie, Y. Dai, R. Elbadry, V. Jani, X. Peng, L. Qian, G. Georgiev, D. Dimitrov, F. Zhang, J. Huang, J. Geng, Y. Chen, Y. Yuan, H. Wu, Y. Wang, I. Koychev, V. Stoyanov, M. Song, Y. Chen, S. Liu, P. Nakov, Overview of FinMMEval 2026: Multilingual and multimodal financial evaluation, in: Experimental IR Meets Multilinguality, Multimodality, and Interaction,...

    2026

  3. [3]

    Z. Xie, L. Qian, G. Georgiev, D. Dimitrov, R. Elbadry, F. Zhang, X. Peng, J. Huang, V. Jani, Y. Dai, J. Geng, Y. Chen, Y. Yuan, H. Wu, Y. Wang, I. Koychev, V. Stoyanov, M. Song, Y. Chen, S. Liu, P. Nakov, Overview of the FinMMEval 2026 task 3: Financial decision making, in: CLEF 2026 Working Notes, CEUR Workshop Proceedings, CEUR-WS.org, Jena, Germany, 2026

    2026

  4. [4]

    L. Qian, X. Peng, Y. Wang, V. J. Zhang, H. He, H. Smith, Y. Han, Y. He, H. Li, Y. Cao, Y. Yu, A. Lopez- Lira, P. Lu, J.-Y. Nie, G. Xiong, J. Huang, S. Ananiadou, When agents trade: Live multi-market trading benchmark for LLM agents, 2025.arXiv:2510.11695

    arXiv 2025

  5. [5]

    Moody, L

    J. Moody, L. Wu, Y. Liao, M. Saffell, Performance functions and reinforcement learning for trading systems and portfolios, Journal of Forecasting 17 (1998) 441–470

    1998

  6. [6]

    Jiang, D

    Z. Jiang, D. Xu, J. Liang, A deep reinforcement learning framework for the financial portfolio management problem, 2017.arXiv:1706.10059

    arXiv 2017

  7. [7]

    V. Mnih, A. P. Badia, M. Mirza, A. Graves, T. Lillicrap, T. Harley, D. Silver, K. Kavukcuoglu, Asynchronous methods for deep reinforcement learning, in: Proceedings of the 33rd International Conference on Machine Learning (ICML), volume 48 ofProceedings of Machine Learning Research, PMLR, New York, NY, USA, 2016, pp. 1928–1937

    2016

  8. [8]

    Schulman, F

    J. Schulman, F. Wolski, P. Dhariwal, A. Radford, O. Klimov, Proximal policy optimization algorithms, 2017.arXiv:1707.06347

    arXiv 2017

Show all 29 references
  1. [9]

    T. P. Lillicrap, J. J. Hunt, A. Pritzel, N. Heess, T. Erez, Y. Tassa, D. Silver, D. Wierstra, Continuous control with deep reinforcement learning, in: Proceedings of the 4th International Conference on Learning Representations (ICLR 2016), San Juan, Puerto Rico, 2016

    2016

  2. [10]

    X.-Y. Liu, Z. Xiong, S. Zhong, H. Yang, A. Walid, Practical deep reinforcement learning approach for stock trading, in: Proceedings of the NeurIPS 2018 Workshop on Challenges and Opportunities for AI in Financial Services, Montréal, Canada, 2018

    2018

  3. [11]

    H. Yang, X. Liu, S. Zhong, A. Walid, Deep reinforcement learning for automated stock trading: An ensemble strategy, in: Proceedings of the ACM International Conference on AI in Finance (ICAIF), New York, NY, USA, 2020

    2020

  4. [12]

    A. Ye, J. Xu, V. Veedgav, Y. Wang, Y. Yu, D. Yan, R. Chen, V. Chaudhary, S. Xu, Learning the market: Sentiment-based ensemble trading agents, 2024.arXiv:2402.01441

    2024 arXiv

  5. [13]

    S. Du, H. Shen, A stock prediction method based on deep reinforcement learning and sentiment analysis, Applied Sciences 14 (2024). URL: https://www.mdpi.com/2076-3417/14/19/8747. doi: 10. 3390/app14198747

    2024

  6. [14]

    Araci, Finbert: Financial sentiment analysis with pre-trained language models, 2019

    D. Araci, Finbert: Financial sentiment analysis with pre-trained language models, 2019. arXiv:1908.10063

    2019 arXiv

  7. [15]

    X. Yu, Z. Chen, Y. Lu, Harnessing LLMs for temporal data – a study on explainable financial time series forecasting, in: Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing (EMNLP): Industry Track, Singapore, 2023, pp. 739–753

    2023

  8. [16]

    Lopez-Lira, Y

    A. Lopez-Lira, Y. Tang, Can ChatGPT forecast stock price movements? Return predictability and large language models, 2025.arXiv:2304.07619

    2025

  9. [17]

    R. S. Sutton, A. G. Barto, Reinforcement Learning: An Introduction, second ed., The MIT Press, 2018

    2018

  10. [18]

    Ferrell, financial-news-multisource (revision b509ef6), 2025

    B. Ferrell, financial-news-multisource (revision b509ef6), 2025. doi:10.57967/hf/6432

    2025 doi

  11. [19]

    K. R. French, Stock returns and the weekend effect, Journal of Financial Economics 8 (1980) 55–69

    1980

  12. [20]

    J. J. Murphy, Technical Analysis of the Financial Markets, New York Institute of Finance, 1999

    1999

  13. [21]

    Brock, J

    W. Brock, J. Lakonishok, B. LeBaron, Simple technical trading rules and the stochastic properties of stock returns, Journal of Finance 47 (1992) 1731–1764

    1992

  14. [22]

    J. W. Wilder, New Concepts in Technical Trading Systems, Trend Research, 1978

    1978

  15. [23]

    Appel, Technical Analysis: Power Tools for Active Investors, Financial Times Prentice Hall, 2005

    G. Appel, Technical Analysis: Power Tools for Active Investors, Financial Times Prentice Hall, 2005

    2005

  16. [24]

    Bollinger, Bollinger on Bollinger Bands, McGraw-Hill, 2001

    J. Bollinger, Bollinger on Bollinger Bands, McGraw-Hill, 2001

    2001

  17. [25]

    A. W. Lo, H. Mamaysky, J. Wang, Foundations of technical analysis: Computational algorithms, statistical inference, and empirical implementation, Journal of Finance 55 (2000) 1705–1765

    2000

  18. [26]

    Grattafiori, A

    A. Grattafiori, A. Dubey, A. Jauhri, A. Pandey, et al., The LLaMA 3 herd of models (2024). arXiv:2407.21783

    2024 arXiv

  19. [27]

    A. Y. Ng, D. Harada, S. J. Russell, Policy invariance under reward transformations: Theory and application to reward shaping, in: Proceedings of the Sixteenth International Conference on Machine Learning (ICML), Morgan Kaufmann, 1999, pp. 278–287

    1999

  20. [28]

    R. J. Williams, Simple statistical gradient-following algorithms for connectionist reinforcement learning, Machine Learning 8 (1992) 229–256

    1992

  21. [29]

    van Hasselt, A

    H. van Hasselt, A. Guez, D. Silver, Deep reinforcement learning with double Q-learning, in: Proceedings of the Thirtieth AAAI Conference on Artificial Intelligence (AAAI-16), Association for the Advancement of Artificial Intelligence, Phoenix, Arizona, 2016, p. 2094–2100

    2016

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.