ARGUS Pipeline Audits Climate-Policy Causal Evidence with LLMs
September 28, 2026
ARGUS uses a structured language-model pipeline to audit difference-in-differences studies against an eleven-dimension rubric. The system detected 73% of planted flaws in benchmarks, significantly outperforming keyword-based methods, and abstains when retrieval evidence is insufficient.
HOW THIS AFFECTS YOU
●
researcherYou can use this to identify weaknesses in the identification assumptions of your causal evaluations.
●
policyYou can automate the auditing of causal claims in environmental policy research.