FinRiskAtlas: Decision-Aligned Evaluation of Large Language Models for Financial Risk Review

Recent research highlights the application of Large Language Models in specialized fields. FinRiskAtlas introduces a Chinese-language benchmark designed to evaluate financial LLMs based on operation execution and decision-aligned evidence for risk review. Simultaneously, a new LLM pipeline was developed to automate systematic literature reviews of disease spread models. Testing on 536 papers showed GPT-5.0 achieved 81.67% paper-level accuracy, though performance varied across complex fields when compared to human-conducted reviews.