FinCARDS: Card-Based Analyst Reranking for Financial Document Question Answering

ACL ARR 2026 January Submission572 Authors

Keywords: financial question answering, evidence reranking, structured retrieval, zero-shot LLMs, auditability

Abstract: Financial question answering (QA) over long corporate filings requires evidence to satisfy strict constraints on entities, financial metrics, fiscal periods, and numeric values. However, existing LLM-based rerankers primarily optimize semantic relevance, leading to unstable rankings and opaque decisions on long documents. We propose FinCards, a structured reranking framework that reframes financial evidence selection as constraint satisfaction under a finance-aware schema. FinCards represents filing chunks and questions using aligned schema fields (entities, metrics, periods, and numeric spans), enabling deterministic field-level matching. Evidence is selected via a multi-stage tournament reranking protocol with stability-aware aggregation, producing auditable decision traces. Across two corporate filing QA benchmarks, FinCards substantially improves early-rank retrieval quality over both lexical and LLM-based reranking baselines, while significantly reducing ranking variance, without requiring model fine-tuning or unpredictable inference budgets. The code is available at https://anonymous.4open.science/r/Fincards-0414.

Paper Type: Long

Research Area: Financial Applications and Time Series

Research Area Keywords: financial question answering, evidence reranking, structured retrieval, zero-shot LLMs, auditability

Contribution Types: Model analysis & interpretability, NLP engineering experiment

Languages Studied: English

Submission Number: 572