{"paper_id":"2510.05702","title":"Uncovering Representation Bias for Investment Decisions in Open-Source Large Language Models","authors":["Fabrizio Dimino","Krati Saxena","Bhaskarjit Sarmah","Stefano Pasquali"],"paper_date":"2025-10-07","paper_type":"software-tool","paper_url":"https://arxiv.org/pdf/2510.05702v2","page_url":"https://thequant.space/flowcharts/uncovering-representation-bias-for-investment-decisions-in-o/","scores":{"empirical_rigor":7,"math_complexity":5.5,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper proposes a balanced round-robin prompting method with constrained decoding and token‑logit aggregation to generate firm‑level confidence scores for about 150 U.S. equities across financial contexts.","quote":"We propose a balanced round-robin prompting method over approximately 150 U.S. equities, applying constrained decoding and token–logit aggregation to derive firm-level confidence scores across financial contexts.","summary":"Large Language Models are increasingly adopted in financial applications to support investment workflows. However, prior studies have seldom examined how these models reflect biases related to firm size, sector, or financial characteristics, which can significantly impact decision-making. This…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"The analysis is limited to a sample of roughly 150 U.S. firms from 2017 to 2024 with pre‑specified feature sets.","type":"data","stated":"explicit","quote":"∼150 U.S. firms (2017–2024) and pre‑specified feature sets; findings may vary with broader universes, alternative sampling, or non‑U.S. markets."},{"assumption":"Correlations reported are descriptive and not interpreted as causal relationships.","type":"statistical","stated":"explicit","quote":"Correlations are descriptive rather than causal, and we do not evaluate out-of-sample trading performance."},{"assumption":"No out‑of‑sample trading performance is evaluated, implying that any implied trading results would ignore execution costs.","type":"execution","stated":"implied","quote":"we do not evaluate out-of-sample trading performance."}]},"datasets":{"status":"available","items":[{"name":"U.S. equities sample","provider":null,"kind":"fundamentals","access":"unknown","period":"2017–2024","universe":"~150 U.S. equities","quote":"approximately 150 U.S. equities"}],"known":[],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"weekly-monthly"},"transaction_costs":{"status":"not_applicable","modeled":"n/a","headline_net_of_costs":null,"detail":null,"quote":null,"source":"front_matter","rule":"R3","candidates":{"front_matter":"n/a","evidence_card":"n/a","evidence_card_quoted":true},"card_detail":null,"card_quote":"we do not evaluate out-of-sample trading performance.","meaning":"the paper reports no trading or execution result"},"results":{"status":"available","evaluation_type":"unclear","findings":[{"finding":"Firm size and valuation consistently increase model confidence, while risk factors tend to decrease it.","kind":"statistical","quote":"we find that firm size and valuation consistently increase model confidence, while risk factors tend to decrease it."},{"finding":"Confidence varies significantly across sectors, with the Technology sector showing the greatest variability.","kind":"statistical","quote":"Confidence varies significantly across sectors, with the Technology sector showing the greatest variability."},{"finding":"When prompted for specific financial categories, LLM confidence rankings align best with fundamental data (free cash flow), moderately with technical signals, and least with growth indicators.","kind":"statistical","quote":"their confidence rankings best align with fundamental data, moderately with technical signals, and least with growth indicators."},{"finding":"Free cash flow shows the strongest positive correlation with LLM ranking preferences across all model sizes (e.g., 0.519*** for Qwen2.5‑32B).","kind":"statistical","quote":"LLM ranking preferences align most strongly and consistently with free cash flow, showing significant positive correlations across all model…"},{"finding":"Smaller models (e.g., Qwen2.5‑7B) tend to anchor more tightly to sectoral ordering than larger counterparts.","kind":"statistical","quote":"smaller models (e.g., Qwen3-8B, Qwen2.5-7B) tend to anchor more tightly than larger counterparts."}],"reported_performance":null,"out_of_sample":{"method":"none","headline_out_of_sample":null}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":35,"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":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":"2026-11-14T17:25:17Z"},"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/"}