YouRA paper targets evidence-traceable autonomous research agents
Tags AI / ML
A team at Korea's Electronics and Telecommunications Research Institute posted YouRA, a persistent-state architecture for autonomous research agents that maintains research state, execution evidence and failure history across long-horizon pipelines. YouRA combines a Verification State Architecture tracking hypotheses, gates and evidence pointers; an Independent Controller separating control from execution; and Stateful Reflection that logs failures as structured lessons. On MLR-Bench's predefined ten-task end-to-end subset, YouRA improved on both MLR-Agent and AI Scientist V2 on scalar Overall across three matched backbones, and ablations showed each core component contributes. The work addresses a structural gap in which manuscript claims diverge from executed experiments.
Technical significance
Treating claim-evidence alignment as persistent, verifiable state addresses a known failure mode of end-to-end research agents, where generated papers overstate what was actually run. The ablation results suggest the state-tracking and control-separation components, not just model scale, drive the improvement.