{"paper_id":"2407.18324","title":"AMA-LSTM: Pioneering Robust and Fair Financial Audio Analysis for Stock Volatility Prediction","authors":["Shengkun Wang","Taoran Ji","Jianfeng He","Mariam Almutairi","Dan Wang","Linhan Wang","Min Zhang","Chang-Tien Lu"],"paper_date":"2024-07-03","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2407.18324v1","page_url":"https://thequant.space/flowcharts/ama-lstm-pioneering-robust-and-fair-financial-audio-analysi/","scores":{"empirical_rigor":7,"math_complexity":5.5,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper proposes an adversarial multimodal attentive LSTM that uses adversarial training to mitigate randomness and gender bias in stock volatility prediction from earnings call audio.","quote":"We suggest a method of adversarial training that tackles the issues of randomness and bias, and implement it on a deep learning model for stock volatility prediction.","summary":"Stock volatility prediction is an important task in the financial industry. Recent advancements in multimodal methodologies, which integrate both textual and auditory data, have demonstrated significant improvements in this domain, such as earnings calls (Earnings calls are public available and…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"Financial audio data contains stochastic (random) information that harms model robustness.","type":"data","stated":"explicit","quote":"We delve into the stochastic nature of financial data, emphasizing the critical need to address the randomness inherent in input features."},{"assumption":"Earnings call audio exhibits gender bias that can affect volatility predictions.","type":"data","stated":"explicit","quote":"We study the impact of gender bias in earnings call audio on stock volatility predictions by training on gender-specific audio features."},{"assumption":"Transformer models require more data to learn their many parameters effectively.","type":"model","stated":"explicit","quote":"The Transformer struggles with temporal information and lacks enough data to learn its many parameters effectively."},{"assumption":"Adversarial perturbations can simulate market noise and bias during training.","type":"model","stated":"explicit","quote":"This method employs adversarial training to more effectively simulate the market noise and biases during model training."}]},"datasets":{"status":"available","items":[{"name":"ACL19","provider":null,"kind":"benchmark","access":"public","period":null,"universe":null,"quote":"Table 3 displays SMA-LSTM’s performance across two datasets."},{"name":"CIKM20","provider":null,"kind":"benchmark","access":"public","period":null,"universe":null,"quote":"Table 3 displays SMA-LSTM’s performance across two datasets."}],"known":[],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"daily"},"transaction_costs":{"status":"not_applicable","modeled":"n/a","headline_net_of_costs":null,"detail":null,"quote":null,"source":"front_matter","rule":"R1","candidates":{"front_matter":"n/a","evidence_card":"unclear","evidence_card_quoted":false},"meaning":"the paper reports no trading or execution result"},"results":{"status":"available","evaluation_type":"forecast-accuracy","findings":[{"finding":"AMA-LSTM achieves the best MSE in almost all cases, beating the Transformer‑based M3ANet by 20.83% and 13.95% across the two datasets.","kind":"performance","quote":"Our method achieves the best MSE in almost all cases. It surpasses the previous state-of-the-art Transformer-based methods and LSTM‑based approaches. Specifically, AMA-LSTM beats Transformer-based M3ANet by 20.83% and 13.95% across two datasets."},{"finding":"Adversarial training reduces gender bias, yielding the lowest ∆MSE compared to models without adversarial training.","kind":"other","quote":"We evaluate the gender bias by comparing ∆MSE, showing that AMA-LSTM achieves the best results in both short-term and long-term across two datasets compared to models without adversarial training."},{"finding":"Adversarial perturbations outperform random stochastic perturbations, improving MSE by 16.05% on ACL19 and 23.71% on CIKM20.","kind":"performance","quote":"Specifically, adversarial training outperforms stochastic perturbation by 16.05% on the ACL19 and 23.71% on the CIKM."}],"reported_performance":null,"out_of_sample":{"method":"holdout","headline_out_of_sample":null}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":18,"citations":{"cited_by_in_corpus":3,"by_independent_authors":2,"how_cited":{"used_as_baseline":0,"uses_method_or_data":0,"extends":0,"discusses_findings":0,"mentions":2,"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":"2026-11-14T18:06:46Z"},"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/"}