# Fighting Numerical Hallucinations via Data-centric Compilation for Online Financial QA

arXiv: https://arxiv.org/abs/2605.31064 · HTML: https://mlsanity.com/paper/2605.31064/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 22 of its 30 references that are in the map.
- First posted on arXiv 29 May 2026.

## Builds on (most cited first, 22 of 22)

- [Qwen3 Technical Report](https://mlsanity.com/paper/2505.09388/) (2025)
- [DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning](https://mlsanity.com/paper/2501.12948/) (2025)
- [Large Language Models are Zero-Shot Reasoners](https://mlsanity.com/paper/2205.11916/) (2022)
- [Qwen Technical Report](https://mlsanity.com/paper/2309.16609/) (2023)
- [Kimi k1.5: Scaling Reinforcement Learning with LLMs](https://mlsanity.com/paper/2501.12599/) (2025)
- [Program of Thoughts Prompting: Disentangling Computation from Reasoning for Numerical Reasoning Tasks](https://mlsanity.com/paper/2211.12588/) (2022)
- [FinQA: A Dataset of Numerical Reasoning over Financial Data](https://mlsanity.com/paper/2109.00122/) (2021)
- [TAT-QA: A Question Answering Benchmark on a Hybrid of Tabular and Textual Content in Finance](https://mlsanity.com/paper/2105.07624/) (2021)
- [RAGTruth: A Hallucination Corpus for Developing Trustworthy Retrieval-Augmented Language Models](https://mlsanity.com/paper/2401.00396/) (2023)
- [ConvFinQA: Exploring the Chain of Numerical Reasoning in Conversational Finance Question Answering](https://mlsanity.com/paper/2210.03849/) (2022)
- [A Survey of Large Language Model Agents for Question Answering](https://mlsanity.com/paper/2503.19213/) (2025)
- [Exploring the Numerical Reasoning Capabilities of Language Models: A Comprehensive Analysis on Tabular Data](https://mlsanity.com/paper/2311.02216/) (2023)
- [FinanceReasoning: Benchmarking Financial Numerical Reasoning More Credible, Comprehensive and Challenging](https://mlsanity.com/paper/2506.05828/) (2025)
- [Exposing Numeracy Gaps: A Benchmark to Evaluate Fundamental Numerical Abilities in Large Language Models](https://mlsanity.com/paper/2502.11075/) (2025)
- [Enhancing Financial Question Answering with a Multi-Agent Reflection Framework](https://mlsanity.com/paper/2410.21741/) (2024)
- [UQABench: Evaluating User Embedding for Prompting LLMs in Personalized Question Answering](https://mlsanity.com/paper/2502.19178/) (2025)
- [APOLLO: An Optimized Training Approach for Long-form Numerical Reasoning](https://mlsanity.com/paper/2212.07249/) (2022)
- [Fine-tuning Smaller Language Models for Question Answering over Financial Documents](https://mlsanity.com/paper/2408.12337/) (2024)
- [Zero-Shot Question Answering over Financial Documents using Large Language Models](https://mlsanity.com/paper/2311.14722/) (2023)
- [Program of Thoughts for Financial Reasoning: Leveraging Dynamic In-Context Examples and Generative Retrieval](https://mlsanity.com/paper/2510.13157/) (2025)
- [Enhancing Numerical Reasoning with the Guidance of Reliable Reasoning Processes](https://mlsanity.com/paper/2402.10654/) (2024)
- [Data-Driven Function Calling Improvements in Large Language Model for Online Financial QA](https://mlsanity.com/paper/2604.05387/) (2026)

## Abstract (arXiv)

Large Language Models (LLMs) have significantly advanced online data services, particularly in the domain of financial question answering (FinQA). However, such systems remain susceptible to numerical reasoning hallucinations, which critically undermine reliability in high-stakes financial applications. Although retrieval-augmented generation (RAG) has been widely adopted to ground responses in external knowledge, it introduces three persistent challenges: noise sensitivity, calculation fragility, and an auditability crisis. Existing model-centric approaches, which primarily focus on optimizing either the retriever or generator in isolation, still struggle to address these issues in an integrated manner. In this work, we pioneer a data-centric paradigm and propose a novel framework, the Data-centric Reasoning Compiler (DCRC). The framework operates through three cohesive phases: (1) adversarial data construction, which synthesizes training examples with controlled noise to teach robustness; (2) multi-stage training that cultivates a Data-centric Structuring Agent (DSA) capable of explicit evidence auditing and program synthesis; and (3) a compile-and-execute inference process, where the DSA transforms user queries and retrieved documents into verifiable, executable reasoning programs. This data-driven framework ensures faithful numerical reasoning by design. We conduct extensive experiments on established offline benchmarks and further validate our framework through deployment in a real-world online financial QA system.