跳到正文
arXiv cs.CL· Hyunji Lee, Justin Chih-Yao Chen, Joykirat Singh, Zaid Khan, Elias Stengel-Eskin, Mohit Bansal·· 10 小时前AI 评分40

MINTEval:评估长程智能体系统在多目标干扰下的记忆能力

MINTEval: Evaluating Memory under Multi-Target Interference in Long-Horizon Agent Systems

AI 导读

研究者提出 MINTEval 基准,用于评估记忆增强智能体在长程、高干扰场景下的表现,包含 15.6k 问答对,上下文平均 138.8k tokens、最长 1.8M tokens,覆盖状态追踪、多轮对话、Wikipedia 修订与 GitHub 提交四类领域。

正文 · 原文

View PDF HTML (experimental)

Abstract:Real-world agents operate over long and evolving horizons, where information is repeatedly updated and may interfere across memories, requiring accurate recall and aggregated reasoning over multiple pieces of information. However, existing benchmarks focus on static, independent recall and fail to capture these dynamic interactions between evolving memories. In this paper, we study how current memory-augmented agents perform in realistic, interference-heavy, long-horizon settings across diverse domains and question types. We introduce MINTEval (Long-Horizon Memory under INTerference Evaluation), a benchmark featuring (1) long, highly interconnected contexts with frequently updated information that induces substantial interference, (2) diverse domains (state tracking, multi-turn dialogue, Wikipedia revisions, and GitHub commits), enabling evaluation of domain generalization, and (3) diverse question types that assess robustness to interference, including (i) single-target recall tasks requiring retrieval of a specific target from long contexts, and (ii) multi-target aggregation tasks requiring reasoning over multiple relevant pieces of information. Overall, MINTEval has 15.6k question-answering pairs over long-horizon contexts averaging 138.8k tokens and extending up to 1.8M tokens per instance. We evaluate 7 representative systems, including vanilla long-context LLMs, RAG, and memory-augmented agent frameworks. Across all systems, we observe consistently low performance (avg. 27.9% accuracy), especially on questions requiring aggregated reasoning over multiple pieces of evidence. Our analysis shows that performance is primarily limited by retrieval and memory construction. Furthermore, current memory systems struggle to recall and reason over earlier facts that are revised or interfered with by subsequent context, with accuracy degrading as the number of intervening updates increases.
Comments: NeurIPS 2026 (Camera-Ready); Equal contribution; order decided by a coin flip. Code and data: this https URL
Subjects: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
Cite as: arXiv:2605.18565 [cs.CL]
  (or arXiv:2605.18565v3 [cs.CL] for this version)
  https://doi.org/10.48550/arXiv.2605.18565

arXiv-issued DOI via DataCite

Submission history

From: Hyunji Lee [view email]
[v1] Mon, 18 May 2026 15:43:35 UTC (332 KB)
[v2] Tue, 19 May 2026 16:05:55 UTC (332 KB)
[v3] Thu, 19 Nov 2026 18:16:11 UTC (313 KB)

来源:arXiv cs.CL · arxiv.org