multi-agent news, stories and analysis

Bar chart comparing structural match rates across five agent architectures on CDISC-Bench ADSL: DAG at 100%, Keyword-RAG at 59.2%, SingleAgent at 0%, FlatMulti at 0%, Embedding-RAG at 0%

LLM agents hit 100% on clinical trial coding with DAG

GxP-Agent encodes regulatory process steps as a directed acyclic graph, letting Claude Sonnet produce valid CDISC analysis datasets where every single-shot LLM attempt scored 0%.

Lars Cornelissen · Aug 19, 2026

Bar chart comparing routing gain across multi-agent configurations. Specialized route shows 94.2% accuracy, random route drops to 61.8%, and single-best-agent baseline sits at 78.3%, showing the gap RouteGuard certifies.

AI

RouteGuard certifies when multi-agent routing actually helps

Multi-agent LLM routing fails when agents overlap too much. RouteGuard is a certification method that proves routing gain before you deploy, catching complementarity gaps that benchmarks miss.

Lars Cornelissen · Aug 11, 2026

Multi-agent debate effects in data cleaning: generation down 1.6 to 15.5 points, detection up 27.4 points.

Research

Multi-agent debate needs a boring data-cleaning cop

Multi-agent debate hurt generation by up to 15.5 points in data cleaning, but a grounded critic rescued detection and repair.

Lars Cornelissen · Jun 3, 2026