LLM pipelines for financial extraction work across languages without retraining

Open-source LLMs tested for automated ESG reporting and KPI extraction

HQ arXiv: AI + Accounting arXiv · 1mo ago · 1 min read

Researchers evaluate open-source large language models with retrieval-augmented generation for automating ESG KPI extraction from 498 real-world EU corporate reports, addressing a gap in domain-spe...

Stanford releases open SEC filings dataset for LLM pretraining

HQ arXiv: AI + Accounting arXiv · 4mo ago · 1 min read

Stanford researchers publish SEFD, an open-source dataset reconstructing SEC EDGAR filings into layout-faithful MultiMarkdown to address scarcity of high-quality long-context training data for fina...

Google's Gemini 4 Argon tops accounting AI benchmark at 91.1% accuracy

HQ Dual Entry Blog blogscraper · 7d ago · 1 min read

Google's Gemini 4 Argon achieved 91.1% accuracy and ranked #1 across 101 accounting tasks on the DualEntry benchmark, demonstrating state-of-the-art performance in accounting-specific AI evaluation.

Study reveals LLMs struggle to align financial concepts across languages

HQ arXiv: AI + Accounting arXiv · 8d ago · 1 min read

Academic research benchmarks multilingual financial evidence extraction, showing language models can recognize accounting concepts across languages but fail to consistently recover all numerical fa...

Study evaluates confidence scoring for text-to-SQL financial queries

arXiv: AI + Accounting arXiv · 10d ago · 1 min read

Researchers audit confidence scoring mechanisms in conversational SQL systems used for database analytics, testing reliability on 1,534 questions across BIRD and Spider benchmarks to ensure trustwo...