Researchers have developed a system called Reinforcement Learning from Community Feedback (RLCF) that enables AI to learn scientific taste. Scientific taste refers to the ability of experts to judge and propose research ideas with long-term impact. The RLCF system consists of two components: Scientific Judge, which learns from community feedback such as citations, and Scientific Thinker, which proposes research ideas with high potential impact. Experiments showed that AI using RLCF outperformed strong language model baselines in judging the quality of research papers and proposing new ideas. This breakthrough suggests that AI can learn to accelerate scientific discovery by reducing reliance on human experts. --- Why it matters: This matters because it could enable AI systems to augment human researchers, accelerating the pace of scientific discovery and potentially leading to breakthroughs in various fields. Source: https://arxiv.org/abs/2603.14473