{"paper_id":"2404.07452","title":"RiskLabs: Predicting Financial Risk Using Large Language Model based on Multimodal and Multi-Sources Data","authors":["Yupeng Cao","Zhi Chen","Prashant Kumar","Qingyun Pei","Yangyang Yu","Haohang Li","Fabrizio Dimino","Lorenzo Ausiello","K. P. Subbalakshmi","Papa Momar Ndiaye"],"paper_date":"2024-04-11","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2404.07452v2","page_url":"https://thequant.space/flowcharts/risklabs-predicting-financial-risk-using-large-language-mod/","scores":{"empirical_rigor":7,"math_complexity":6,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper proposes the RiskLabs framework, a multimodal LLM-based system for financial risk prediction.","quote":"we propose the RiskLabs framework, designed to comprehensively understand the complexities of the investment environment.","summary":"The integration of Artificial Intelligence (AI) techniques, particularly large language models (LLMs), in finance has garnered increasing academic attention. Despite progress, existing studies predominantly focus on tasks like financial text summarization, question-answering, and stock movement…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"The prior distribution of model parameters follows a Normal distribution.","type":"statistical","stated":"explicit","quote":"We assume that the prior distribution 𝑃(θ) satisfies the Normal distribution based on the simple linear regression."},{"assumption":"The independent variable is independent of the regression coefficient.","type":"model","stated":"explicit","quote":"We also suppose that the independent variable is independent with the regression coefficient;"}]},"datasets":{"status":"available","items":[{"name":"Earnings Conference Call Transcripts and Audio","provider":null,"kind":"text-news","access":"unknown","period":null,"universe":null,"quote":"The multimodal data sources include: (1) earnings conference call transcripts, offering insights into corporate performance and future prospects; (2) audio from these calls, capturing tone and sentiment not evident in the text; (3) time series data for histor…"}],"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":"RiskLabs outperforms other models in prediction accuracy, especially for short‑term and medium‑term forecasts, achieving the lowest MSE values.","kind":"performance","quote":"RiskLabs framework outperforms others in prediction accuracy, particularly in short-term and medium-term forecasts, evidenced by the lowest Mean Squared Error (MSE) values."},{"finding":"The Audio + Text combination beats the baseline HTML model for 3‑day forecasts.","kind":"predictive-accuracy","quote":"As shown in Table 2, the Audio + Text’ combination outperforms the baseline HTML model for 3-day forecasts."},{"finding":"Integrating earnings‑call analysis and time‑series data yields incremental gains for medium‑ to long‑term forecasts (7, 15, 30 days).","kind":"performance","quote":"Further improvements were observed upon integrating earnings conference call analysis and time-series data, leading to incremental gains in RiskLabs’ performance, especially for medium- to long-term forecasts (7, 15, and 30 days)."},{"finding":"RiskLabs demonstrates high efficacy in predicting financial risks, confirming its value as a tool.","kind":"predictive-accuracy","quote":"The RiskLabs framework demonstrates a high efficacy in predicting financial risks, confirming its potential as a valuable tool in this domain."},{"finding":"Ablation studies show each module (Earnings Call Encoder, Time‑Series Encoder, News‑Market Reactions Encoder) meaningfully contributes to final risk prediction accuracy.","kind":"statistical","quote":"Ablation studies further affirm that each individual module of RiskLabs meaningfully contributes to the accuracy of the final risk predictions."}],"reported_performance":null,"out_of_sample":{"method":"unclear","headline_out_of_sample":null}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":55,"citations":{"cited_by_in_corpus":6,"by_independent_authors":5,"how_cited":{"used_as_baseline":0,"uses_method_or_data":0,"extends":0,"discusses_findings":0,"mentions":5,"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":"2026-11-14T16:35:31Z"},"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/"}