{"paper_id":"2407.15038","title":"Explainable AI in Request-for-Quote","authors":["Qiqin Zhou"],"paper_date":"2024-07-21","paper_type":"methodological","paper_url":"https://arxiv.org/pdf/2407.15038v1","page_url":"https://thequant.space/flowcharts/explainable-ai-in-request-for-quote/","scores":{"empirical_rigor":4,"math_complexity":7,"quadrant":"Lab Rats"},"proposes":{"status":"available","statement":"The paper proposes a new hybrid model called Bayesian Neural Tree (BNT) that combines decision tree explainability with neural network predictive power, and a market‑maker utility function to optimize quote prices.","quote":"To balance the trade-off between model performance and model explainability, we propose a new model framework: Bayesian Neural Tree (BNT).","summary":"In the contemporary financial landscape, accurately predicting the probability of filling a Request-For-Quote (RFQ) is crucial for improving market efficiency for less liquid asset classes. This paper explores the application of explainable AI (XAI) models to forecast the likelihood of RFQ…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"Side is a binary variable following a Bernoulli distribution with probability 0.5.","type":"data","stated":"explicit","quote":"Side is binary variable and we assume it follows a a Bernoulli distribution with probability p = 0.5."},{"assumption":"There are exactly 4 participants in each RFQ, with CounterParty drawn uniformly from [0,3].","type":"data","stated":"explicit","quote":"We assume there are in all 4 participants in RFQ so CounterParty is generated uniformly from [0, 3]."},{"assumption":"Competition (number of competitors) is uniformly distributed between 1 and 4.","type":"data","stated":"explicit","quote":"Competition is generated uniformly from [1, 4] since there are at most 4 participants in the RFQ process."},{"assumption":"Status (RFQ filled or missed) follows a Bernoulli distribution independent of Side.","type":"data","stated":"explicit","quote":"Status is the target of our prediction and is also a binary variable. We assume it follows a Bernoulli distribution that is independent of Side."},{"assumption":"Bid‑ask spread evolves according to a Geometric Brownian Motion.","type":"model","stated":"explicit","quote":"The spread at each time step t then can be updated using a Geometric Brownian Motion model: St+1 = St exp( µ∆t + σSϵt √∆t )."},{"assumption":"Mid‑price updates use stochastic processes with normal shocks for bid and ask.","type":"model","stated":"explicit","quote":"Bt+1 = Mt −St+1 2 + ϵB At+1 = Mt + St+1 2 + ϵA where ϵB ∼N(0, σ2 B) and ϵA ∼N(0, σ2 A) are random shocks drawn from normal distributions."}]},"datasets":{"status":"available","items":[{"name":"synthetic RFQ dataset","provider":null,"kind":"synthetic","access":"unknown","period":null,"universe":null,"quote":"In Section III, we first proposed a simulation algorithm to produce hypothetical RFQ data that are otherwise not available to the public because of the regulatory constraints that limit the sharing and distribution of real RFQ data."}],"known":[],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"synthetic"},"transaction_costs":{"status":"not_applicable","modeled":"n/a","headline_net_of_costs":null,"detail":null,"quote":null,"source":"evidence_card","rule":"R2","candidates":{"front_matter":"unclear","evidence_card":"n/a","evidence_card_quoted":false},"meaning":"the paper reports no trading or execution result"},"results":{"status":"available","evaluation_type":"forecast-accuracy","findings":[{"finding":"The Bayesian Neural Network model achieves about 90% accuracy on out‑of‑sample data.","kind":"performance","quote":"The Bayesian Neural Network model accuracy on out of sample dataset is around 90%."},{"finding":"Ensembling multiple models (Lasso Logistic, Random Forest, XGBoost, Bayesian Neural Network) improves out‑of‑sample accuracy and mitigates over‑fitting.","kind":"performance","quote":"Ensembling successfully mitigates the problem overfitting and improve accuracy on out-of-sample data."},{"finding":"The most important features for predicting RFQ fill are response, MOM5, MOM20 and counterparty.","kind":"predictive-accuracy","quote":"As we can see in figure 5, the most importance features are response, MOM5, MOM20 and counterparty."},{"finding":"When simulating quote prices, the model fills 4 out of 5 out‑of‑sample RFQs and secures positive utility for all.","kind":"performance","quote":"Among all 5 RFQs out of sample, we are able to fill 4 RFQ and secure positive utility values on all 5 of them."},{"finding":"Bayesian Neural Tree balances good predictive performance with strong explainability due to its hierarchical tree structure.","kind":"theoretical","quote":"We found that Bayesian Neural Tree has balanced off a good performance and a strong explainability with its hierarchical tree structure."}],"reported_performance":null,"out_of_sample":{"method":"holdout","headline_out_of_sample":null}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":34,"citations":{"cited_by_in_corpus":0,"by_independent_authors":0,"how_cited":{"used_as_baseline":0,"uses_method_or_data":0,"extends":0,"discusses_findings":0,"mentions":0,"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":"2026-11-14T18:05:49Z"},"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/"}