Which Agent's Skill Should Change? Group-Relative Credit Assignment for Evolving Multi-Agent Teams

Key points

• Cooperative language-model agents can improve without weight updates by revising reusable skills, but only if a shared team reward is traced to the responsible agent and skill.1
• Existing methods credit the skills present in each trajectory, yet presence is not responsibility: a harmful skill active in successful runs still earns positive credit.1
• On HotpotQA and LoCoMo with two open-weight actors, GRSE outperforms MASkills, the closest prior method, on HotpotQA and every LoCoMo category except open-domain, with the largest gains on multi-hop questions (+6.7 and +3.2 F1).1
• GRSE is robust to corrupted libraries: starting from planted harmful skills, it credits every planted skill with the correct sign and recovers the score of a clean library (+6.3 points), whereas MASkills misattributes two and falls below its starting point.1

Sources (2)

• [1]Which Agent's Skill Should Change? Group-Relative Credit Assignment for Evolving Multi-Agent Teams
  Hugging Face Daily Papers · Oct 19, 12:00 AM

  We present GRSE, Group-Relative Skill Evolution, which assigns behavioral credit by comparison: inspired by GRPO, it executes each question several times under the same skill libraries and contrasts the executions to credit agent--skill pairs, with structured diagnosis of failures.

  Cooperative language-model agents can improve without weight updates by revising reusable skills, but only if a shared team reward is traced to the responsible agent and skill.

• [2]Which Agent's Skill Should Change? Group-Relative Credit Assignment for Evolving Multi-Agent Teams
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 19, 07:55 PM · same content

Extractive summary: sentences quoted from the sources.