{"paper_id":"2306.07013","title":"Combining Reinforcement Learning and Barrier Functions for Adaptive Risk Management in Portfolio Optimization","authors":["Zhenglong Li","Hejun Huang","Vincent Tam"],"paper_date":"2023-06-12","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2306.07013v1","page_url":"https://thequant.space/flowcharts/combining-reinforcement-learning-and-barrier-functions-for-a/","scores":{"empirical_rigor":7,"math_complexity":8.5,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper proposes the RiPO framework, the first integration of reinforcement learning and barrier function based constraint programming for portfolio optimization.","quote":"The RiPO framework is the first attempt to integrate RL and BF-based constraint programming for financial applications.","summary":"Reinforcement learning (RL) based investment strategies have been widely adopted in portfolio management (PM) in recent years. Nevertheless, most RL-based approaches may often emphasize on pursuing returns while ignoring the risks of the underlying trading strategies that may potentially lead to…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"The portfolio will be only considered from long positions.","type":"model","stated":"explicit","quote":"Assumption 1. The portfolio will be only considered from long positions in this work."},{"assumption":"The turnover rate of assets in a portfolio satisfies the requirements of each order execution.","type":"execution","stated":"explicit","quote":"Assumption 2. The turnover rate of assets in a portfolio satisfies the requirements of each order execution."}]},"datasets":{"status":"available","items":[{"name":"S&P500 constituent stocks daily data","provider":"Yahoo Finance","kind":"market-prices","access":"public","period":"2015-2022 (approx.)","universe":"Top 10 S&P500 stocks by market cap","quote":"the daily OLHCV data of constitute stocks of S&P500 index in the U.S. market is collected from Yahoo Finance. The top 10 stocks are selected to construct a portfolio","known_id":"yahoo-finance"}],"known":[{"id":"yahoo-finance","name":"Yahoo Finance","provider":"Yahoo","kind":"market-prices","access":"public","quote":"5.1 Experimental Settings Datasets: To evaluate the performance of methods in the real finan-cial market, the daily OLHCV data of constitute stocks of S&P500 index in the U.S. market is collected from Yahoo Finance.","mentions":1,"used_in":1}],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"daily"},"transaction_costs":{"status":"available","modeled":"yes","headline_net_of_costs":null,"detail":"Transaction cost ς and slippage ξ (uniformly sampled) are included in the reward function.","quote":"To close the realistic trading environment, two practical factors including transaction cost ς and slippage ξ are considered in each transaction. The reward at timestamp t can be defined as rt = [−ς +","source":"front_matter","rule":"R3","candidates":{"front_matter":"yes","evidence_card":"yes","evidence_card_quoted":true},"card_detail":"Transaction cost ς and slippage ξ (uniformly sampled) are included in the reward function.","card_quote":"To close the realistic trading environment, two practical factors including transaction cost ς and slippage ξ are considered in each transaction. The reward at timestamp t can be defined as rt = [−ς +","meaning":"costs are part of the reported headline results"},"results":{"status":"available","evaluation_type":"trading-pnl","findings":[{"finding":"RiPO achieves returns at least 5% higher than other methods, with a maximum drawdown around 22% and the best Sharpe Ratio of 0.72, demonstrating a balance of profit and risk.","kind":"performance","quote":"the returns of the RiPO framework are at least 5% higher than other methods while performing relatively low MDD at around 22%"},{"finding":"Increasing the impact of the risk controller (higher m) improves AR from –9.43% to –6.58% and reduces MDD from 29.47% to 25.56% in downtrend markets, though overly high m may miss profitable opportunities.","kind":"performance","quote":"the higher m performs better AR (from −9.43% to −6.58%) and lower MDD (from 29.47% to 25.56%) by enhancing the impact of risk controller to the RL-based agents in downtrend markets."},{"finding":"Adaptive Risk Strategy (ARS) reduces losses from 23.32% to 6.69% (without DCM) and from 18.76% to 6.58% (with DCM), while also halving risk exposure.","kind":"performance","quote":"the ARS significantly reduces the losses from 23.32% to 6.69% when without using the DCM and from 18.76% to 6.58% when integrating the DCM. Meanwhile, the risk exposure decreases to half of the setting without the ARS."},{"finding":"Using a scaling factor µ=3 in MS-2 helps RiPO balance expected returns and short-term risks by dynamically adjusting risk constraint strength.","kind":"performance","quote":"using the appropriate scaling factor µ at 3 in the MS-2 encourages the RiPO framework to balance the expected returns and short-term risks by dynamically adjusting the strength of risk constraints."}],"reported_performance":null,"out_of_sample":{"method":"holdout","headline_out_of_sample":null}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":38,"citations":{"cited_by_in_corpus":2,"by_independent_authors":2,"how_cited":{"used_as_baseline":0,"uses_method_or_data":0,"extends":0,"discusses_findings":0,"mentions":1,"unclassified":1},"independent_papers":[],"scope":"Citations between papers in this corpus only; citations from outside it are not counted."}},"implementation":{"status":"available","public":false,"code_url":null,"note":"No repository presented by the authors as their own was found in the paper. Absence here is not proof that no code exists."},"replication":{"status":"none_found","card":null,"note":"No replication of this paper has been run by thequant.space yet."},"extraction":{"text_source":"fulltext","model":"openai/gpt-oss-120b","version":1,"extracted_at":"2026-11-14T18:25:55Z"},"provenance":{"claims":"Everything here is what the authors state in the paper. Nothing was re-run or reproduced by the platform.","quotes":"Assumptions, datasets, cost detail and findings were extracted by a language model. Every quote is verbatim paper text. Where an item has `support`, \"full\" means every figure in it appears in the quoted sentences and \"partial\" means a figure is stated elsewhere in the paper; items with a figure found nowhere in the paper were discarded.","datasets_known":"datasets.known lists well-known vendors and public datasets by a stable id, found by a dictionary over the full text where the sentence says the data is used; it is independent of the model and covers only names in the dictionary.","flags":"transaction_costs.modeled, results.evaluation_type and results.out_of_sample.method are platform classifications (structured-decision models); see https://thequant.space/score-guide/#penalty-flags","comparisons":"supports / contradicts / qualifies are a language model's reading of two papers' quoted findings, restricted to closely related papers that share no author. They are pointers to read, not verified replications, and coverage is partial.","implementation":"code_url is a repository the paper itself presents as its own (rule-scored, ambiguous cases judged by a model), checked for liveness when it was recorded."},"docs":"https://thequant.space/agents/"}