AutoSFT splits SFT data research across multi-agent roles
The result suggests automated data research depends on action diversity across roles rather than one stronger agent — a design lesson for anyone building research agents. Practitioners running post-training pipelines can try the multi-role decomposition, and code plus paper are public. The single-agent baseline's context hitting 94.7% and compressing 10 times is a useful stress marker for long-horizon agent designs.
