Oblivion: Self-Adaptive Agentic Memory Control through Decay-Driven Activation

2026-03-31

AI Summary

This work addresses the challenges of severe memory interference and high latency in existing memory-augmented large language model agents during long-term interactions, which stem from always-on retrieval and flat memory storage. To overcome these limitations, the authors propose Oblivion—a memory control framework inspired by human selective forgetting—that models forgetting as a decay in memory accessibility rather than explicit deletion. Oblivion introduces a hierarchical memory organization through decoupled read and write pathways: the read path dynamically determines when to query external memory based on uncertainty and buffer sufficiency, while the write path reinforces critical memories according to their contribution to responses, guided by a decay-driven activation mechanism. Experiments demonstrate that Oblivion effectively balances learning and forgetting in both static and dynamic long-horizon tasks, significantly enhancing agent reasoning efficiency and performance in evolving contexts.

Abstract

Human memory adapts through selective forgetting: experiences become less accessible over time but can be reactivated by reinforcement or contextual cues. In contrast, memory-augmented LLM agents rely on "always-on" retrieval and "flat" memory storage, causing high interference and latency as histories grow. We introduce Oblivion, a memory control framework that casts forgetting as decay-driven reductions in accessibility, not explicit deletion. Oblivion decouples memory control into read and write paths. The read path decides when to consult memory, based on agent uncertainty and memory buffer sufficiency, avoiding redundant always-on access. The write path decides what to strengthen, by reinforcing memories contributing to forming the response. Together, this enables hierarchical memory organization that maintains persistent high-level strategies while dynamically loading details as needed. We evaluate on both static and dynamic long-horizon interaction benchmarks. Results show that Oblivion dynamically adapts memory access and reinforcement, balancing learning and forgetting under shifting contexts, highlighting that memory control is essential for effective LLM-agentic reasoning. The source code is available at https://github.com/nec-research/oblivion.

Problem

• memory-augmented LLM agents
• always-on retrieval
• flat memory storage
• memory interference
• latency

Innovation

• memory control
• selective forgetting
• decay-driven activation
• hierarchical memory
• LLM agents