Failure-Gated Hierarchical Memory: Preventing Memory Pollution in Long-Horizon LLM Agents via Critic-Guided Write Control
Open accessThe DOI matches its Crossref record; 24 of the 24 references it cites were found in a registry or an index.
Abstract
Large language model (LLM)-based agents equipped with persistent memory can accumulate experience across tasks, yet uncontrolled memory writes introduce memory pollution—erroneous, outdated, or over-generalized knowledge entries that silently degrade downstream task performance through retrievalaugmented generation. We present Failure-Gated Hierarchical Memory (FGHM), a three-tier memory architecture (working memory, episodic buffer, procedural longterm memory) augmented with a dedicated Failure Critic agent that gates all writes from the episodic buffer to procedural longterm memory. The Critic, implemented as a LoRA-tuned Llama-3-8B model trained on hard-negative knowledge candidates derived from failure trajectories, issues APPROVE, REJECT_REDUNDANT, or REJECT_CONFLICT verdicts for each candidate entry.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Citations
Cites
LoRA: Low-Rank Adaptation of Large Language Models
Hu, E. J., Shen, Y., Wallis, P., Allen-Zhu, Z., Li, Y., Wang, S., Wang, L., & Chen, W. (2021) · arXiv · doi:10.48550/arxiv.2106.09685 · Cited by 13,453
Human Memory: A Proposed System and its Control Processes
Atkinson, R. C., & Shiffrin, R. M. (1968) · Psychology of Learning and Motivation · doi:10.1016/s0079-7421(08)60422-3 · Cited by 7,216 · cited 2×
“Our memory hierarchy is inspired by the Atkinson-Shiffrin multi-store model [19] and Tulving’s distinction between episodic and semantic memory [20].”
(2007) · Cognitive Psychology · doi:10.4135/9781446212967.n15 · Cited by 3,749
Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks
Lewis, P., Perez, E., Piktus, A., Petroni, F., Karpukhin, V., Goyal, N., Küttler, H., Lewis, M., Yih, W.-T., Rocktäschel, T., et al. (2020) · arXiv · doi:10.48550/arxiv.2005.11401 · Cited by 3,101 · cited 1×
“The agent maintains a memory system M that persists across episodes, influencing future behavior through retrieval-augmented generation[23].”
A survey on large language model based autonomous agents
Wang, L., Ma, C., Feng, X., Zhang, Z., Yang, H., Zhang, J., Chen, Z., Tang, J., Chen, X., Lin, Y., et al. (2024) · Frontiers of Computer Science · doi:10.1007/s11704-024-40231-1 · Cited by 1,941
Expel: LLM agents are experiential learners
Zhao, A., Huang, D., Xu, Q., Lin, M., Liu, Y.-J., & Huang, G. (2024) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v38i17.29936 · Cited by 1,451
Reflexion: Language Agents with Verbal Reinforcement Learning
Shinn, N., Cassano, F., Berman, E., Gopinath, A., Narasimhan, K., & Yao, S. (2023) · arXiv · doi:10.48550/arxiv.2303.11366 · Cited by 1,087 · cited 4×
“FIFO buffers discard entries indiscriminately, MemoryBank [11] relies on decay rather than correctness, and Reflexion [9] stores verbal self-critiques without independent verification.”
InfluentialCognitive Architectures for Language Agents
Sumers, T. R., Yao, S., Narasimhan, K., & Griffiths, T. L. (2023) · arXiv · doi:10.48550/arxiv.2309.02427 · Cited by 798
Direct Preference Optimization: Your Language Model is Secretly a Reward Model
Rafailov, R., Sharma, A., Mitchell, E., Manning, C. D., Ermon, S., & Finn, C. (2023) · Advances in Neural Information Processing Systems 36 · doi:10.52202/075280-2338 · Cited by 686
MemoryBank: Enhancing Large Language Models with Long-Term Memory
Zhong, W., Guo, L., Gao, Q., Ye, H., & Wang, Y. (2023) · arXiv · doi:10.48550/arxiv.2305.10250 · Cited by 253
Research works printed but not linked: 3
- Chhikara, 2025, Mem0: Building production-ready AI agent memory, arXiv preprint arXiv
- Ebbinghaus, Über das Gedächtnis: Untersuchungen zur experimentellen Psychologie
- 2024, Llama 3 model card
Metadata
- Authors
- Liu, Q., Wang, M., & Xie, J.
- Affiliations
- National Science Library (CN)
- ORCID
- Wang, Meng
- Venue
- 2026 6th International Conference on Machine Learning and Intelligent Systems Engineering (MLISE)
- Publisher
- IEEE
- Published
- 2026-05
- Pages
- 132-139
- OpenAlex
- W7170034571
- Cited by
- 0
- References
- 14
- FWCI
- 0.00
- Type
- conference paper
- Language
- en
- Open access
- hybrid
- doi.org
- Licence
- Locations
- International Conference on Machine Learning and Intelligent Systems Engineering (MLISE) (publishedVersion, cc-by)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › AI-based Problem Solving and Planning
- Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling
- Physical Sciences › Computer Science › Artificial Intelligence › Reinforcement Learning in Robotics
- Keywords
- LLM agents, hierarchical memory, long-term memory, episodic buffer, retrieval-augmented generation, ALFWorld, WebShop
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: Failure-Gated Hierarchical Memory: Preventing Memory Pollution in Long-Horizon LLM Agents via Critic-Guided Write Control. https://app.cochat.ai/papers/10.1109/mlise70044.2026.11607496
- The paper
- Liu, Q., Wang, M., & Xie, J. (2026). Failure-Gated Hierarchical Memory: Preventing Memory Pollution in Long-Horizon LLM Agents via Critic-Guided Write Control. 2026 6th International Conference on Machine Learning and Intelligent Systems Engineering (MLISE). https://doi.org/10.1109/mlise70044.2026.11607496
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.