Automating ML for Science: Can Frontier Agents Climb Scientific Hills in the Wild?
What this work does
AGENT4SCI provides 25 tasks from 20 scientific benchmarks across 11 domains. It measures how coding agents improve verified baselines under fixed resources and investigates whether those improvements involve domain insight.
THE RESEARCH CONVERSATION
Discussion & reactions
Share a question, an insight, or a replication experience. Sign in with GitHub to join the conversation.