Author of "AIRS-Bench: A Suite of Tasks for Frontier AI Research Science Agents" on arXiv
How media typically covers Despoina Magka
Despoina Magka as author
AIRS-Bench introduces a 20-task benchmark for evaluating AI research agents across the full scientific lifecycle, showing agents exceed human performance on 4 tasks but fail on 16 others with substantial room for improvement.
“Author of "AIRS-Bench: A Suite of Tasks for Frontier AI Research Science Agents" on arXiv”