🛰️ Daily AI Frontier
‹ back to 2026-08-26

CAFE: Self-Improving Search Agents Need Co-Evolving Feedback

Research LLM Agents

Ranking

Overall 83
Content 95
Popularity 56

Observed public metrics from 1 member.

Representative image for CAFE: Self-Improving Search Agents Need Co-Evolving Feedback

Merged summary

TL;DR - CAFE jointly improves an agentic search policy and its feedback critic by training them in alternating, coupled roles. Across seven in-domain and six out-of-domain benchmarks, this approach outperformed evaluated RL-based search agents on average and reduced answer-level hallucinations.

  • A shared-parameter model alternates between acting as the search agent and providing corrective, in-trajectory feedback.
  • Online RL rewards useful feedback requests using a call-versus-skip success gap and reshapes token-level advantages around feedback.
  • Offline preference optimization trains feedback from matched successful and unsuccessful rollout trajectories.
  • Ablations found that improving only the agent or critic eventually plateaued, while alternating updates continued to improve performance.

Sources (1)

CAFE: Self-Improving Search Agents Need Co-Evolving Feedback

arXiv cs.AI Boyang Liu, Senjie Jin, Peixin Wang, Zhangyue Yin, Yibo Wang, Yuhao Zhou, Xinbing Liang, Shizheng Zhu, Yuhui Wang, Jingqi Tong, Zhiheng Xi, Jiazheng Zhang, Clive Bai, Clarenceai, Blaze Chen, Tao Gui, Qi Zhang, Xuanjing Huang 2026-08-25 arXiv:2608.24794
Public signals Hugging Face upvotes 6 · Semantic Scholar citations 0 · Semantic Scholar influential citations 0
Providers: Hugging Face · Upvotes 6 OpenAlex · N/A Publisher · N/A Semantic Scholar · Citations 0 · Influential citations 0 X · N/A Fetched 2026-09-25 14:28:40.119365 UTC

TL;DR - CAFE jointly improves an agentic search policy and its feedback critic by training them in alternating, coupled roles. Across seven in-domain and six out-of-domain benchmarks, this approach outperformed evaluated RL-based search agents on average and reduced answer-level hallucinations.

  • A shared-parameter model alternates between acting as the search agent and providing corrective, in-trajectory feedback.
  • Online RL rewards useful feedback requests using a call-versus-skip success gap and reshapes token-level advantages around feedback.
  • Offline preference optimization trains feedback from matched successful and unsuccessful rollout trajectories.
  • Ablations found that improving only the agent or critic eventually plateaued, while alternating updates continued to improve performance.
item →