Agent memory improves event combining for better question answers
Beyond Retrieval: Evidence Access and Event Aggregation in Agent Memory
Artificial Intelligence
Summary
It's often hard for AI agents to answer questions about their past because just looking up events is not enough—they need to add up or combine information from multiple past events. The authors created new methods that help the AI remember exactly where information came from and how to combine multiple pieces in a reliable way. Their system, TallyMem, improves how well AI answers questions about past code changes by nearly 14 percentage points compared to previous methods. This shows that knowing how to gather and add up events helps agents understand and explain their history better.
What this means in practice
- •For software development teams: Improve automated tools that answer questions about code changes by accurately combining multiple events from version histories.
- •For automated customer support teams: Enhance systems that need to summarize and combine multiple past interactions to provide better answers to user questions.
Authors
Zerui Yang, Quan Xue, Chaorui Zhang, Xueyan Niu, Xuanyu Zheng, Xiyuan Wang, Haolin Ruan, Guangcheng Song, Weixi Zhang, Wei Han, Bo Bai
Abstract
Questions about an agent's history often require aggregating events, which retrieval alone does not resolve. We combine source-preserving evidence access with executable aggregation. Trajectory Memory (TM) retains the original trajectory and constructs a source-addressable reading view at question time. TallyMem augments it with a reusable event ledger, model-generated event specifications and deterministic execution. With a commit-specific adaptation on 776 questions from 99 held-out MINTEval repositories, TallyMem raises Qwen3-32B exact match from 35.2% to 49.0%, a 13.8-point gain over TM; gains span 12.8-13.9 points across three readers. A controlled probe with event sets parsed independently from raw logs shows an 11.0-point gain, from 82.8% to 93.8%, without routing. On full AMA-Bench, the routed system yields a smaller 1.3-point gain, from 69.8% to 71.1%. These results support executable aggregation on tasks that require combining events, with benefits dependent on the task and event specification, the largest residual system-error category in our diagnostic analysis.