The paper introduces MERIT, a benchmark that evaluates the marginal benefit of long‑term memory for tool‑using large language model agents while explicitly accounting for cost. MERIT provides episodic tool‑use tasks across three domains, verifies dependence on earlier‑episode facts, and measures memory operations in tokens and dollars. Experiments on GPT‑4.1‑mini, Claude Haiku 4.5, and Claude Sonnet 5 show that memory can significantly improve task success, but its utility varies widely across models and memory implementations, and full replay is rarely cost‑effective.
By Shweta Mishra, Shashank Mishra
arXiv:2607. 08032v1 Announce Type: new Abstract: Large language models, and the agents built on them, spend an ever-growing share of their compute and memory on remembering: caching attention keys and values, carrying long prompts, maintaining recurrent state, and storing what happened in previous turns and sessions.
By Ashwin Gerard Colaco, Nada Lahjouji
MemCut introduces a benchmark to test user-requested forgetting in large language model agents, featuring 456 hand-designed secrets embedded in long-horizon conversations with fully annotated evidence trails. The benchmark defines two recovery metrics—native and residual—to assess how well secrets can be retrieved after deletion. Experiments across five memory designs and four LLMs show that deletion reduces native recovery more than residual recovery, highlighting candidate coverage as a key bottleneck.
By Cheng Zhang, Xuanning Cao, Zhicheng Zheng, Hui Xiong
arXiv:2605. 18421v2 Announce Type: replace-cross Abstract: Recent benchmarks for Large Language Model (LLM) agents mainly evaluate reasoning, planning, and execution.
By Yuyao Wang, Zhongjian Zhang, Mo Chi, Kaichi Yu, Yuhan Li, Miao Peng, Bing Tong, Chen Zhang, Yan Zhou, Jia Li
arXiv:2609. 04875v1 Announce Type: cross Abstract: Long-running LLM agents are stateful: beyond the transcript they accrete compressed summaries, plaintext memory, pending tool plans, and, under every serving API, a KV cache.
By Chao Yao, Yangbo Wei, Zhen Huang, Junhong Qian, Chenle Chen, Shaoqiang Lu, Chen Wu, Lei He
arXiv:2607. 01071v1 Announce Type: cross Abstract: Memory has emerged as a cornerstone of modern LLM-based agents, supporting their evolution from single-turn assistants to long-term collaborators.
By Zhishang Xiang, Zerui Chen, Yunbo Tang, Zhimin Wei, Ruqin Ning, Yujie Lin, Qinggang Zhang, Jinsong Su
The paper investigates how large language model agents can over-rely on agentic memory, a phenomenon where retrieved memories distort inference even when they are correctly stored and retrieved. It shows that memory is helpful when past experience fully transfers to the current task but becomes misleading under partial query-memory overlap, a pattern confirmed by controlled experiments. To address this, the authors propose MEMTRIM, a plug‑and‑play framework that indexes memory evidence at write time and limits its reuse at read time, thereby reducing over-reliance without retraining and preserving useful memory benefits across models and memory architectures.
By Luoxi Tang, Yuqiao Meng, Nilesh Auradkar, Muchao Ye, Dazheng Zhang, Zhaohan Xi
MemoryArena is a new evaluation gym that benchmarks agent memory in interdependent multi‑session tasks. Unlike prior benchmarks that test memorization or single‑session action in isolation, MemoryArena requires agents to acquire memory while interacting with the environment and then use that memory to guide future decisions across a range of tasks such as web navigation, planning, information search, and formal reasoning. The benchmark reveals that agents excelling on existing long‑context memory tests perform poorly here, highlighting a gap in current memory evaluation methods.
By Zexue He, Yu Wang, Churan Zhi, Yuanzhe Hu, Tzu-Ping Chen, Lang Yin, Ze Chen, Tong Arthur Wu, Siru Ouyang, Zihan Wang, Jiaxin Pei, Julian McAuley, Yejin Choi, Alex Pentland
Memory has emerged as a cornerstone of modern LLM-based agents, supporting their evolution from single-turn assistants to long-term collaborators. However, memory is not always beneficial: retrieved memories often induce a critical issue of sycophancy, causing agents to over-align with the user at the cost of factual accuracy or objective reasoning.
DyadMem introduces a new benchmark for evaluating how long‑term agents remember user‑specific relational information, called User‑conditioned Relational Agent Memory (URAM). The dataset contains 3,065 episodes, 50,961 sessions, and 61,210 QA pairs, with detailed annotations for memory capture, update, and recall across multi‑session trajectories. Experiments on 20 models show strong performance in a gold‑memory setting but a significant drop in full‑pipeline QA, highlighting gaps in current LLM memory capabilities.
By Yifei Tao, Xinyu Zhong, Henry Hengyuan Zhao, Fanyi Wang, Tengda Guo, Wentao Qiu, Ying Wang, Liujian Tang
arXiv:2609.09115v1 Announce Type: new
Abstract: Long horizon Large Language Model (LLM) agents rely on external memory systems to preserve user preferences and task knowledge across extended interact...
By Boyu Yang, Jiazheng Sun, Zilong Lu, Zhi Qiu, Xin Peng, Jun Zheng
arXiv:2507. 05257v4 Announce Type: replace-cross Abstract: Recent benchmarks for Large Language Model (LLM) agents primarily focus on evaluating reasoning, planning, and execution capabilities, while another critical component-memory, encompassing how agents memorize, update, and retrieve long-term information-is under-evaluated due to the lack of benchmarks.
By Yuanzhe Hu, Yu Wang, Julian McAuley