{"paper_id":"2507.18417","title":"FinDPO: Financial Sentiment Analysis for Algorithmic Trading through Preference Optimization of LLMs","authors":["Giorgos Iacovides","Wuyang Zhou","Danilo Mandic"],"paper_date":"2025-07-24","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2507.18417v1","page_url":"https://thequant.space/flowcharts/findpo-financial-sentiment-analysis-for-algorithmic-trading/","scores":{"empirical_rigor":8.5,"math_complexity":3,"quadrant":"Street Traders"},"proposes":{"status":"available","statement":"The paper proposes FinDPO, a finance‑specific LLM framework aligned with human preferences via Direct Preference Optimization for financial sentiment analysis and trading.","quote":"We introduce FinDPO, the first finance-specific LLM framework based on post-training human preference alignment via Direct Preference Optimization (DPO).","summary":"Opinions expressed in online finance-related textual data are having an increasingly profound impact on trading decisions and market movements. This trend highlights the vital role of sentiment analysis as a tool for quantifying the nature and strength of such opinions. With the rapid development…","strategy":"FinDPO sentiment-based long-short portfolio"},"assumptions":{"status":"available","items":[{"assumption":"Transaction costs are assumed to lie between 1 and 5 basis points, typical for U.S. equities.","type":"execution","stated":"explicit","quote":"we varied the transaction cost parameter 𝑘from 1 to 5 basis points (bps), which aligns with the typical 2–5 bps range observed in U.S. equities."},{"assumption":"The investable universe consists of liquid S&P 500 constituent stocks.","type":"market","stated":"explicit","quote":"Given that our investable universe consisted of relatively liquid instruments, specifically the S&P 500 constituents"},{"assumption":"A daily rebalancing, long‑short strategy with equal long and short position sizes is used.","type":"execution","stated":"explicit","quote":"Given that our investable universe consisted of relatively liquid instruments, specifically the S&P 500 constituents, we varied the transaction cost parameter…"},{"assumption":"Risk‑free rate is set to 0% for performance calculations.","type":"statistical","stated":"explicit","quote":"a 0% rate is commonly used and was adopted in our analysis."},{"assumption":"Preference pairs are generated by prompting a frozen reference LLM to produce dispreferred responses.","type":"model","stated":"explicit","quote":"we prompted the reference model, 𝜋ref, with ˜𝑥𝑖 to generate sentiment predictions."}]},"datasets":{"status":"available","items":[{"name":"Financial PhraseBank (FPB) Dataset","provider":"HuggingFace","kind":"text-news","access":"public","period":null,"universe":null,"quote":"The FPB dataset consists of 4,840 samples which were randomly extracted from financial news articles.","known_id":"financial-phrasebank"},{"name":"Twitter Financial News Sentiment (TFNS) Dataset","provider":"HuggingFace","kind":"text-news","access":"public","period":null,"universe":null,"quote":"The TFNS dataset includes 11,930 tweets with content from the financial domain.","known_id":"twitter"},{"name":"GPT‑labelled Financial News (NWGI) Dataset","provider":"HuggingFace","kind":"text-news","access":"public","period":null,"universe":null,"quote":"The NWGI dataset consists of 16,200 financial news articles."},{"name":"S&P 500 price data","provider":null,"kind":"market-prices","access":"public","period":null,"universe":"S&P 500 constituents","quote":"Given that our investable universe consisted of relatively liquid instruments, specifically the S&P 500 constituents"}],"known":[{"id":"financial-phrasebank","name":"Financial PhraseBank","provider":"Malo et al.","kind":"benchmark","access":"public","quote":"The training data consisted of three pub-licly available labeled financial news datasets, obtained from Hug-gingFace: the Financial PhraseBank (FPB) dataset [17], the Twitter Financial News dataset [25], and the GPT-labeled Financial News dataset [16].","mentions":2,"used_in":2},{"id":"twitter","name":"Twitter / X","provider":"X Corp.","kind":"text-news","access":"public","quote":"Each sample was annotated with one of the three labels: positive, negative, and neutral. • Twitter Financial News Sentiment (TFNS).","mentions":2,"used_in":2}],"sample":{"start":"2015","end":"2021","market":"S&P 500 stocks","asset_class":"equities","frequency":"daily"},"data_resolution":"daily"},"transaction_costs":{"status":"available","modeled":"unclear","headline_net_of_costs":true,"detail":null,"quote":null,"source":"conflict","rule":"R4b","candidates":{"front_matter":"yes","evidence_card":"partial","evidence_card_quoted":true},"conflict_kind":"graded","card_detail":"Costs varied from 1 to 5 basis points (bps) and were applied in a robustness check; headline results are reported without costs.","card_quote":"we varied the transaction cost parameter 𝑘from 1 to 5 basis points (bps), which aligns with the typical 2–5 bps range observed in U.S. equities.","meaning":"the platform classifier and the paper-text extraction disagree, so no value is asserted"},"results":{"status":"available","evaluation_type":"trading-pnl","findings":[{"finding":"FinDPO achieves state‑of‑the‑art sentiment classification performance, improving average weighted F1 by 11% over the SOTA FinGPT v3.3 model.","kind":"predictive-accuracy","quote":"our proposed FinDPO model has achieved a weighted F1 score that outperforms the SOTA FinGPT v3.3 model by 11% on the average."},{"finding":"In a frictionless setting FinDPO delivers cumulative return of 747.10% (annualized 111.78%) with Sharpe 3.41, Sortino 6.05 and Calmar 11.94, far surpassing all baselines.","kind":"performance","quote":"FinDPO (Ours) 747.10 111.78 3.41 6.05 11.94"},{"finding":"FinDPO outperforms the prior best sentiment‑based portfolio (FinLlama) by 187% in cumulative returns and improves Sharpe ratio by 1.1.","kind":"performance","quote":"... achieving a 187% improvement in cumulative returns."},{"finding":"Even under realistic transaction costs of up to 5 bps, FinDPO maintains substantial positive returns and strong risk‑adjusted performance.","kind":"performance","quote":"FinDPO is the first sentiment‑based approach to maintain substantial positive returns of 67% annually and strong risk‑adjusted performance, as indicated by a Sharpe ratio of 2.0, even under realistic transaction costs of 5 basis points (bps)."}],"reported_performance":{"strategy":"FinDPO sentiment-based long-short portfolio","sharpe":2,"sharpe_benchmark":0.62,"annual_return_pct":67,"total_return_pct":747.1,"max_drawdown_pct":null},"out_of_sample":{"method":"unclear","headline_out_of_sample":null}},"independent_evidence":{"status":"available","supports":[],"contradicts":[],"qualifies":[],"compared_with":40,"citations":{"cited_by_in_corpus":1,"by_independent_authors":1,"how_cited":{"used_as_baseline":0,"uses_method_or_data":1,"extends":0,"discusses_findings":0,"mentions":0,"unclassified":0},"independent_papers":[{"paper":{"paper_id":"2601.03948","title":"Trade-R1: Bridging Verifiable Rewards to Stochastic Environments via Process-Level Reasoning Verification","paper_date":"2026-01-07","authors":["Rui Sun","Yifan Sun","Sheng Xu","Li Zhao","Jing Li","Daxin Jiang"],"empirical_rigor":6,"page_url":"https://thequant.space/flowcharts/trade-r1-bridging-verifiable-rewards-to-stochastic-environm/"},"how":"uses_method_or_data","quote":null}],"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":"2026-11-14T17:31:33Z"},"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/"}