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 with stability-aware aggregation, producing auditable decision traces. Across two corporate filing QA benchmarks, FINCARDS substantially improves early-rank retrieval over both lexical and LLM-based reranking baselines, while reducing ranking variance, without requiring model fine-tuning or unpredictable inference budgets. Our code is available at https://github.com/XanderZhou2022/FINCARDS.

In Text-to-SQL tasks, large language models can generate structurally different SQL queries that return correct answers for the same intent. We call this phenomenon execution-correct structural divergence (ECSD). We introduce SQLStructEval, a framework that analyzes this behavior through canonical abstract syntax tree representations. We quantify structural diversity and agreement across generations. Experiments with different LLMs on multiple Text-to-SQL datasets, including Spider, document ECSD across models and datasets. Furthermore, our experiments demonstrate that generated queries are sensitive to question paraphrases and schema presentation. To address ECSD, we adopt a pipeline that first generates structured intermediate representations and then deterministically compiles them into SQL, improving execution accuracy and structural agreement among correct outputs. Structural analysis thus provides an additional diagnostic perspective that complements execution-based evaluation. Code is available at https://xanderzhou2022.github.io/AACL2026-SQLSTRUCTEVAL/.

LLM persona control divides between trait-prompt strings and model-specific activation-steering vectors; neither offers a per-response trace from output to persona contents. We introduce Persona-CAPS, a retrieval-based persona framework grounded in the Cognitive-Affective Personality System (Mischel and Shoda, 1995): a persona is a corpus of if (situation) then (behavior) contingencies mined from public Experience Sampling Method data, retrieved per-situation, each fired atom logged to its source observation. An audit-loop study quantifies three structural properties paragraph baselines lack: retrieval exposes corpus-level corruption that paragraph dilution masks (P²-sanity passes 0/9 on MiniMax low-Openness under identical polarity injection); atom-level repair uses ~1/3 the paragraph-baseline character footprint at comparable fix success (byte-identical on 13/13 paired GPT-4o cross-vendor conditions); and two audit modes (universal-recall sweep, sparse-edit triage) support deployment-time governance. Across six backbones, CAPS beats Simple-Prompt on all three backbones of PersonalityBench (Deng et al., 2025) (mean +0.69/10), delivers monotonic 1-D continuous control via mixture weights π on all 15 (vendor × Big-Five) Mini-IPIP cells, and matches the strongest paragraph baseline (P²) on the non-all-low 5-D profiles (§4.6; backbone-level direction-match ≥ 0.93). Cross-family judge replication (r = +0.951) and a two-rater human pack triangulate the headline. Code, corpus, and logs are released.

Financial reporting systems increasingly leverage Large Language Models (LLMs) to extract and summarize corporate disclosures. However, most existing approaches assume a single-market setting and overlook structural differences across jurisdictions. Variations in accounting taxonomies, tagging infrastructures (e.g., XBRL vs. PDF), and aggregation conventions introduce substantial challenges for semantic alignment and reliable verification. Here, we aim to bridge this gap. We present FinReporting, an agentic workflow for localized cross-jurisdiction financial reporting. The system constructs a unified canonical ontology spanning the income statement, balance sheet, and cash flow statement, and decomposes reporting into auditable stages, including filing acquisition, extraction, canonical mapping, and anomaly logging. Rather than treating LLMs as free-form generators, FinReporting employs them as constrained verifiers operating under explicit decision rules with evidence grounding.Evaluated on annual filings from the USA, Japan, and China, FinReporting improves consistency and reliability under heterogeneous reporting regimes. We further release an interactive demo that enables cross-market inspection and supports structured export of localized financial statements. Our demo is available at https://huggingface.co/spaces/BoomQ/FinReporting-Demo. A video describing our system is available at https://www.youtube.com/watch?v=f65jdEL31Kk.