SkillRubric: Co-Evolving Actor Guidance and Evaluator Rubrics for Multimodal Agents.
arXiv:2609.34557v1 Announce Type: new Abstract: Recent work incorporates reusable skills distilled from past interactions into multimodal agent training, providing procedural guidance for long-horizon planning and tool use. However, policy optimization in thes
arXiv cs.AIResearch source·