{"paper_id":"2501.07508","title":"Improving DeFi Accessibility through Efficient Liquidity Provisioning with Deep Reinforcement Learning","authors":["Haonan Xu","Alessio Brini"],"paper_date":"2025-01-13","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2501.07508v1","page_url":"https://thequant.space/flowcharts/improving-defi-accessibility-through-efficient-liquidity-pro/","scores":{"empirical_rigor":8,"math_complexity":7,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper proposes a deep reinforcement learning (DRL) approach using Proximal Policy Optimization to optimize liquidity provisioning in Uniswap v3, enabling an active LP agent to dynamically rebalance its positions.","quote":"a deep reinforcement learning (DRL) approach to optimize liquidity provisioning in Uniswap v3, enabling an LP to act as an agent that systematically re‑balances its position over time.","summary":"This paper applies deep reinforcement learning (DRL) to optimize liquidity provisioning in Uniswap v3, a decentralized finance (DeFi) protocol implementing an automated market maker (AMM) model with concentrated liquidity. We model the liquidity provision task as a Markov Decision Process (MDP) and…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"Liquidity provisioning is modeled as a Markov Decision Process.","type":"model","stated":"explicit","quote":"We model the liquidity provision task as a Markov Decision Process (MDP) and train an active liquidity provider (LP) agent using the Proximal Policy Optimization (PPO) algorithm."},{"assumption":"Hourly discretization of price data is sufficient for decision making.","type":"data","stated":"explicit","quote":"At each discrete hourly step, the agent decides whether to withdraw and redeploy liquidity with adjusted price boundaries or to maintain the current interval, balancing risk and profitability."},{"assumption":"Rolling‑window training captures regime shifts and reflects realistic market conditions.","type":"market","stated":"explicit","quote":"We choose this rolling window approach to reflect the practical challenges of deploying DRL models in real‑world financial markets, particularly in the volatile cryptocurrency market."},{"assumption":"Gas fees are a fixed $5 per deployment action.","type":"execution","stated":"explicit","quote":"where g represents the fixed costs for the gas fees which we set at $5 based on Etherscan’s gas tracker."},{"assumption":"Trading fee rate δ is constant for the pool.","type":"execution","stated":"explicit","quote":"where δ represents the fixed fee rate."},{"assumption":"Liquidity providers face a penalty proportional to instantaneous LVR.","type":"model","stated":"explicit","quote":"... (Milionis et al. 2023) defines instantaneous LVR as ℓt(σ, pt) = Lσ2 /4 √pt"}]},"datasets":{"status":"available","items":[{"name":"Hourly WETH/USDC Uniswap v3 pool data","provider":"Uniswap Ethereum subgraph","kind":"crypto-onchain","access":"public","period":"5 May 2021 1:00 am to 29 January 2024 7:","universe":"WETH/USDC pair at 0.05% fee tier","quote":"The dataset for this study consists of hourly data spanning from 5 May 2021 at 1:00 am, marking the inception of Uniswap v3, to 29 January 2024 at 7:00 pm. We focus on the WETH/USDC Uniswap pool at the 0.05% fee tier.","known_id":"the-graph"}],"known":[],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"intraday-bars"},"transaction_costs":{"status":"available","modeled":"yes","headline_net_of_costs":null,"detail":"Gas fees are modeled as a fixed $5 per position change; trading fees are modeled with a fixed fee rate δ as in equations (5) and (6).","quote":"The reward function includes a penalty for gas fees incurred when deploying liquidity positions","source":"evidence_card","rule":"R2","candidates":{"front_matter":"unclear","evidence_card":"yes","evidence_card_quoted":true},"card_detail":"Gas fees are modeled as a fixed $5 per position change; trading fees are modeled with a fixed fee rate δ as in equations (5) and (6).","card_quote":"The reward function includes a penalty for gas fees incurred when deploying liquidity positions","meaning":"costs are part of the reported headline results"},"results":{"status":"available","evaluation_type":"trading-pnl","findings":[{"finding":"The active LP agent outperforms heuristic passive LP strategies in 7 out of 11 out‑of‑sample windows, achieving higher cumulative rewards.","kind":"performance","quote":"Our empirical results reveal that the active LP out‑performs heuristic‑based passive LP strategies in diverse market conditions, achieving higher cumulative rewards in 7 out of 11 out‑of‑sample windows."},{"finding":"The active LP adapts liquidity intervals dynamically, reducing unnecessary rebalancing costs compared to the passive LP.","kind":"performance","quote":"to maximize fee collection with maintaining inactive positions during anticipated mean‑reversion periods, reducing unnecessary costs."},{"finding":"In stationary price periods the active LP redeploys liquidity only once, while the passive LP’s heuristic rebalancing leads to lower cumulative rewards.","kind":"performance","quote":"In this case, the active LP (bottom‑right panel) redeploys liquidity only once, recognizing the stationary nature of the price. Conversely, the passive LP (bottom‑left panel) adheres to its heuristic rebalancing strategy, which results in unnecessary rebalancing and lower cumulative rewards."}],"reported_performance":null,"out_of_sample":{"method":"walk-forward","headline_out_of_sample":null}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":19,"citations":{"cited_by_in_corpus":1,"by_independent_authors":1,"how_cited":{"used_as_baseline":0,"uses_method_or_data":0,"extends":0,"discusses_findings":0,"mentions":1,"unclassified":0},"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":"2026-11-14T17:48:05Z"},"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/"}