Practitioner brief · model risk & AI governance

Five frontier LLMs read 789 hand-labeled filings paragraphs. The best scored 69%.

A plain-language brief on what a ground-truth benchmark of ChatGPT-4o, Claude, Gemini, Grok, and DeepSeek reveals about LLM accuracy in ESG disclosure classification — and what it means for your model inventory.

Domain model (FinBERT)
83%
Best frontier LLM
69%
accuracy vs. human-annotated ground truth · n = 789 · S&P 100 10-K filings, 2012–2022