#financial-reporting
Generating and auditing financial reports with language models
FinRAGBench-V (EMNLP 2025) is the first large-scale benchmark for multimodal RAG with visual citations in finance, covering 112K+ document pages and 1,394 human-annotated QA pairs. Top models achieve only 20–61% block-level citation recall, and multimodal retrieval outperforms text-only by nearly 50 percentage points.
Fin-RATE benchmarks 17 LLMs on 7,500 expert-curated QA pairs from 2,472 SEC filings, revealing an 18.60% accuracy collapse under longitudinal tracking and a 54-point drop for finance-specialized Fin-R1 on cross-entity tasks — with the retrieval pipeline, not the backbone model, as the binding bottleneck.
FinDER benchmarks RAG on 5,703 real hedge fund analyst queries against S&P 500 10-K filings; E5-Mistral achieves only 25.95% context recall, and abbreviation-heavy queries cost 8.2 precision points — evidence that query normalization, not better embeddings, is the first fix for finance AI pipelines.
DocFinQA replaces FinQA's curated 700-word passages with full 123,000-word SEC filings, exposing a 175× context increase that nearly halves GPT-4 accuracy on long documents. Retrieval pipelines fail to surface the right chunk 45% of the time at HR@3 — and long-context models are not a substitute.
FinAuditing tests 13 LLMs zero-shot on 1,102 real SEC XBRL filing instances; top scores are 13.86% on financial math verification and 12.42% on concept retrieval—results that directly bound what AI accounting tools can be trusted to automate without external tooling.
TAT-LLM fine-tunt LLaMA 2 7B met LoRA op financiële tabel-tekst QA-benchmarks en behaalt 64,60% EM op FinQA — waarmee het de 63,91% van GPT-4 verslaat — door redenering te ontleden in deterministische Extraheer-Redeneer-Voer-uit stappen die rekenkundige fouten elimineren.
MultiHiertt (ACL 2022) introduces 10,440 QA pairs from real financial reports averaging 3.89 hierarchical tables each; state-of-the-art models score 38% F1 versus 87% for humans, with a 15-point penalty for cross-table questions — quantifying the retrieval gap finance AI must close.
ConvFinQA's best model scores 68.9% execution accuracy versus 89.4% for human experts—a 21-point gap that multi-turn ledger chat still faces.
TAT-QA is a 16,552-question benchmark over hybrid table-plus-text financial report contexts that showed evidence grounding — not arithmetic — is the core bottleneck in finance AI; by 2024, fine-tuned 7B LLMs reached 83% F1, closing most of the gap against a 91% human ceiling.
FinQA (EMNLP 2021) built 8,281 QA pairs from S&P 500 earnings reports requiring multi-step arithmetic programs. Neural models scored 61% at release versus 91% for human experts; accuracy collapses to 22% on three-or-more-step programs. The failure modes — domain constants, cross-modality grounding, chain length — map directly to the challenges Beancount agents face today.
FinanceBench evaluates 16 AI configurations against 10,231 questions from real SEC filings; shared-vector-store RAG answers correctly only 19% of the time, and even GPT-4-Turbo with the oracle passage reaches just 85% accuracy — showing that numerical reasoning, not retrieval, is the binding constraint for enterprise finance AI.
PHANTOM (NeurIPS 2025) е първият бенчмарк, който измерва откриването на халюцинации в LLM върху реални SEC документи при дължина на контекста до 30 000 токена. Qwen3-30B-A3B-Thinking води с F1=0.882; моделите от 7B постигат резултати близки до случайното отгатване — с директни последици за автономните счетоводни агенти.
