Recent work has explored improving agents by jointly evolving their harnesses and models, but often takes a ''potpourri'' approach that bundles together new …
机构:复旦
来源:arXiv 2609.35290 | AI4Papers 论文推荐平台