{"paper_id":"2512.23515","title":"Alpha-R1: Alpha Screening with LLM Reasoning via Reinforcement Learning","authors":["Zuoyou Jiang","Li Zhao","Rui Sun","Ruohan Sun","Zhongjian Li","Jing Li","Daxin Jiang","Zuo Bai","Cheng Hua"],"paper_date":"2025-12-29","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2512.23515v1","page_url":"https://thequant.space/flowcharts/alpha-r1-alpha-screening-with-llm-reasoning-via-reinforceme/","scores":{"empirical_rigor":8.5,"math_complexity":6,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper proposes Alpha-R1, an 8B-parameter reasoning model trained via reinforcement learning for context‑aware alpha screening.","quote":"We propose Alpha-R1, an 8B-parameter reasoning model trained via reinforcement learning for context-aware alpha screening.","summary":"Signal decay and regime shifts pose recurring challenges for data-driven investment strategies in non-stationary markets. Conventional time-series and machine learning approaches, which rely primarily on historical correlations, often struggle to generalize when the economic environment changes.…","strategy":"Alpha-R1"},"assumptions":{"status":"available","items":[{"assumption":"Execution assumes a transaction fee of 0.1% per trade to account for commissions and slippage.","type":"execution","stated":"explicit","quote":"A transaction fee of 0.1% (10 bps) is applied to both buy and sell orders to account for commissions and slippage."},{"assumption":"Slot rotation assumes a holding period H (e.g., 5 days) and rebalances only one slot per day, reducing turnover to 1/H.","type":"execution","stated":"explicit","quote":"independent sub‑portfolios (referred to as slots), where"},{"assumption":"VWAP execution assumes price is computed from the first 30 minutes of the trading session as a conservative estimate of slippage.","type":"execution","stated":"explicit","quote":"is computed using transaction data from the first 30 minutes of the trading session (09:31–10:00)."},{"assumption":"The factor backtesting assumes that the historical performance vector Pi (returns, volatility, decay) provides the ground‑truth for factor effectiveness.","type":"data","stated":"explicit","quote":"which includes key metrics such as returns, volatility, and decay characteristics."},{"assumption":"The dynamic factor zoo assumes the Alpha101 library provides 82 computationally feasible factors after filtering.","type":"data","stated":"explicit","quote":"We construct a dynamic factor zoo by filtering the Alpha101 library [18], retaining 82 computationally feasible factors."},{"assumption":"Training assumes the CSI 300 constituent stocks represent a high‑turnover environment for the training phase.","type":"market","stated":"explicit","quote":"Using constituent stocks from the CSI 300, we simulate a high‑turnover environment."}]},"datasets":{"status":"available","items":[{"name":"Alpha101 library","provider":null,"kind":"other","access":"unknown","period":null,"universe":null,"quote":"We construct a dynamic factor zoo by filtering the Alpha101 library [18], retaining 82 computationally feasible factors."},{"name":"CSI 300 constituents","provider":null,"kind":"market-prices","access":"unknown","period":"2024.07.01 – 2024.12.31","universe":"CSI 300","quote":"Using constituent stocks from the CSI 300, we simulate a high‑turnover environment."},{"name":"CSI 1000 universe","provider":null,"kind":"market-prices","access":"unknown","period":null,"universe":"CSI 1000","quote":"The CSI 1000 experiment (Figure 2b) evaluates the model’s ability to generalize under a zero‑shot setting."}],"known":[],"sample":{"start":"2025-01","end":"2025-06","market":"CSI 300 stocks","asset_class":"equities","frequency":"daily"},"data_resolution":"intraday-bars"},"transaction_costs":{"status":"available","modeled":"yes","headline_net_of_costs":true,"detail":"A flat transaction fee of 0.1% (10 bps) per side is applied to both buy and sell orders.","quote":"A transaction fee of 0.1% (10 bps) is applied to both buy and sell orders to account for commissions and slippage.","source":"front_matter","rule":"R3","candidates":{"front_matter":"yes","evidence_card":"yes","evidence_card_quoted":true},"card_detail":"A flat transaction fee of 0.1% (10 bps) per side is applied to both buy and sell orders.","card_quote":"A transaction fee of 0.1% (10 bps) is applied to both buy and sell orders to account for commissions and slippage.","meaning":"costs are part of the reported headline results"},"results":{"status":"available","evaluation_type":"trading-pnl","findings":[{"finding":"Alpha‑R1 achieves a Sharpe ratio of 1.62 and limits maximum drawdown to 6.76% on the CSI 300, outperforming traditional baselines.","kind":"performance","quote":"Alpha‑R1 attains a Sharpe ratio of 1.62 while limiting maximum drawdown to 6.76%."},{"finding":"On the out‑of‑domain CSI 1000 universe, Alpha‑R1 obtains a cumulative return of 42.49% and a Sharpe ratio of 4.03, demonstrating strong zero‑shot generalization.","kind":"performance","quote":"Alpha‑R1 maintains strong performance on the out‑of‑domain CSI 1000 universe, achieving a cumulative return of 42.49% and a Sharpe ratio of 4.03."},{"finding":"Baseline reinforcement‑learning agents (A2C, PPO) suffer negative or marginal returns and low Sharpe ratios in both in‑domain and out‑of‑domain tests.","kind":"negative","quote":"A2C achieves a negative cumulative return of -5.52% with a Sharpe ratio of -0.85, while PPO shows only marginal positive returns (0.89%) with a low Sharpe ratio of 0.11."},{"finding":"Generic reasoning LLMs (Claude 3.7 Sonnet, DeepSeek‑R1) exhibit substantial drawdowns (~15–16%) and negative Sharpe ratios, indicating lack of domain‑specific grounding.","kind":"negative","quote":"They fail to adequately account for risk constraints, leading to substantial drawdowns (approximately 15–16%) and negative Sharpe ratios."},{"finding":"Ablation studies confirm that semantic gating and the integration of news, price, and semantic descriptions are key contributors to Alpha‑R1’s superior performance.","kind":"performance","quote":"tangible advantages over traditional heuristic gating strategies."}],"reported_performance":{"strategy":"Alpha-R1","sharpe":4.03,"sharpe_benchmark":null,"annual_return_pct":78.18,"total_return_pct":42.49,"max_drawdown_pct":6.76},"out_of_sample":{"method":"holdout","headline_out_of_sample":true}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":30,"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":true,"code_url":"https://github.com/FinStep-AI/Alpha-R1","note":null},"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":"2026-11-14T17:17:45Z"},"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/"}