Surrogate-Guided Memory Retrieval for Autonomous Agents
Published in Submitted to the NeurIPS 2026 Workshop on Dynamic Alignment in Human–AI Coupled Systems (BiAlign), 2026
Role: co-first author — proposed and benchmarked the method · Status: submitted to the NeurIPS 2026 Workshop on Dynamic Alignment in Human–AI Coupled Systems (BiAlign).
We study memory retrieval for autonomous agents — deciding which past experiences an agent should recall to act well now. Instead of relying on raw semantic similarity, we train the retriever offline with a learned surrogate that scores how much a candidate memory improves downstream task success. Directly mentored by Prof. Xue (Steve) Liu and Ye Yuan in the Cyber-Physical Intelligence Lab and AAAAA Community.
Recommended citation: Yonghan Yang*, Ye Yuan*, et al. Surrogate-Guided Memory Retrieval for Autonomous Agents. Submitted to the NeurIPS 2026 BiAlign Workshop, 2026. (* equal contribution)