The authors established RSI Arena, a two-stage study of autonomous post-training.
AI agents autonomously train and improve language models from human feedback
Ten frontier models each post-trained a base model; the best agent's model matched a human-designed instruct model in style-controlled human evaluation.
Independent
Yue Huang · Yanbo Wang · Lichi Li · Yuanyuan Chen · Andreas Haupt · Xiaonan Luo · +12 more
Research Digest··3 min read
RSI Arena tasked ten AI agents with autonomously post-training a base language model over six days with a fixed budget, then improving it using human preference data.
The Daily Edition
The front page in your inbox each weekday morning: the lead, the picks, and why each one earned its place.