What to Keep, What to Forget: A Rate--Distortion View of Memory Compaction in LLMs and Agents

• URL: http://arxiv.org/abs/2607.08032v1
• Date: Thu, 09 Jul 2026 01:15:03 GMT
• Title: What to Keep, What to Forget: A Rate--Distortion View of Memory Compaction in LLMs and Agents
• Abstract summary: We show how four separate research communities have each learned to compact large language memory.<n>We argue that these are instances of one problem: a rate--distortion decision about what context-derived information to retain versus discard.<n>We make this lens precise with a single compaction objective and a layer-agnostic lower bound, use it to build a seven-axis taxonomy that classifies methods from across the stack uniformly.
• Score: 1.0742675209112622
• License: http://creativecommons.org/licenses/by/4.0/
• 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. Because none of this memory is free, four largely separate research communities have each learned to compact it. They evict or quantize the KV cache, prune or distill prompts, bound architectural state, and consolidate agent memory. We argue that these are instances of one problem: a rate--distortion decision about what context-derived information to retain versus discard, at what fidelity, under a resource budget, so as to preserve downstream task utility. We make this lens precise with a single compaction objective and a layer-agnostic lower bound, use it to build a seven-axis taxonomy that classifies methods from across the stack uniformly, and use it to transfer mechanisms between layers that have never been connected, from serving-stack KV management to agent long-term memory. Two patterns hold across the survey. At every layer the signal that decides what to keep is attention magnitude or recency, and it fails in the same way everywhere, by discarding, before the query is known and with no way to undo it, information the query later needs. And while compression is measured carefully on single-turn long context, the repeated compaction that agents actually perform is almost never measured, and no benchmark holds one budget axis across all the layers at once. We turn both observations into a benchmark proposal, a small reference experiment, and a set of compaction-aware design principles, and we map the open problems.

Related papers

• Beyond Episodic Evaluation: Memory Architectural Bottlenecks in Sequential Embodied Question Answering [93.80043825360514]
  Embodied question answering (EQA) is traditionally evaluated under an episodic formulation, where agents solve each task independently and reset internal state between episodes.<n>We investigate how different memory architectures behave when EQA agents are evaluated sequentially, with multiple questions answered in the same scene while memory is carried forward across queries.<n>We find that simply preserving existing memory is often insufficient.<n>We highlight the necessity of structured, spatially grounded memory: architectures that map persistent visual observations onto metric 3D geometry preserve visual-semantic evidence in a coherent scene representation.
  arXiv  Detail & Related papers  (2026-07-23T17:50:45Z)
• Memory as a Controlled Process: Learned Adaptive Memory Management for LLM Agents [72.3752981234916]
  Large Language Model (LLM) agents increasingly rely on external memory systems to accumulate experience across tasks.<n>We present Memory as a Controlled Process (MemCon), a framework that models memory operations as a Markov Decision Process.<n>MemCon consistently outperforms multiple memory baselines by up to 15.2 points in task success while reducing token consumption by 5-20%.
  arXiv  Detail & Related papers  (2026-07-15T08:32:40Z)
• MemOps: Benchmarking Lifecycle Memory Operations in Long-Horizon Conversations [50.24315431387575]
  We introduce MemOps, a benchmark that reformulates conversational memory as a sequence of lifecycle operations.<n>MemOps represents each memory event with a structured trace specifying its trigger, target, scope, state transition, and supporting evidence.<n>Across long-context, retrieval-based, parametric and managed-memory systems, MemOps disentangles failure modes that final-answer accuracy alone conceals.
  arXiv  Detail & Related papers  (2026-07-14T15:33:44Z)
• Remembering Distinct Items, Not Tokens: A Learnable Dirichlet-Process Cache Between State-Space Models and Attention [0.48342038441006796]
  We study a sparse cache that allocates a slot only when an input is novel, so its size tracks the number of distinct items rather than the number of tokens.<n>We develop it in two forms, a static cache with a fixed concentration and a surprise-adaptive variant whose concentration follows the recent novelty rate.
  arXiv  Detail & Related papers  (2026-07-10T18:28:21Z)
• WorldMemArena: Evaluating Multimodal Agent Memory Through Action-World Interaction [72.1620416874118]
  Multimodal large language models are increasingly deployed as long-horizon agents.<n>Existing benchmarks measure recall over static dialogue, collapse memory into a single end-of-task accuracy, and reduce visual observations to captions.<n>We formulate multimodal agent memory as an Action-World Interaction Loop with an observable four-stage lifecycle.
  arXiv  Detail & Related papers  (2026-05-28T04:27:20Z)
• Make Each Token Count: Towards Improving Long-Context Performance with KV Cache Eviction [65.710271475739]
  We introduce a global retention-based KV eviction method that learns each token's future utility under a unified memory budget.<n>We show that our method substantially reduces KV memory while matching or surpassing full-cache inference.<n>These results suggest that learned, globally calibrated KV eviction is not only a compression technique, but also a mechanism for improving long-context reasoning.
  arXiv  Detail & Related papers  (2026-05-10T16:47:50Z)
• The Trap of Trajectory: Towards Understanding and Mitigating Spurious Correlations in Agentic Memory [7.707732051915869]
  Diagnosing agentic memory systems reveals that memory improves reasoning on clean inputs but amplifies reliance on spurious patterns when they are present.<n>We propose CAMEL, a plug-and-play calibration method that operates across diverse memory architectures at both write and retrieval time.<n>Overall, CAMEL offers a principled and lightweight solution toward more reliable agentic memory deployment.
  arXiv  Detail & Related papers  (2026-05-10T05:04:13Z)
• Contextual Agentic Memory is a Memo, Not True Memory [3.939241105154204]
  We argue that treating lookup as memory is a category error with provable consequences for agent capability, long-term learning, and security.<n>We show that biological intelligence solved this problem by pairing fast hippocampal storage with slow neocortical weight consolidation.
  arXiv  Detail & Related papers  (2026-04-30T10:54:56Z)
• Memory for Autonomous LLM Agents:Mechanisms, Evaluation, and Emerging Frontiers [0.42061757959666934]
  Large language model (LLM) agents increasingly operate in settings where a single context window is too small to capture what has happened.<n>Memory is what turns a stateless text generator into a genuinely adaptive agent.<n>This survey offers a structured account of how memory is designed, implemented, and evaluated in modern LLM-based agents.
  arXiv  Detail & Related papers  (2026-03-08T15:08:01Z)
• Beyond RAG for Agent Memory: Retrieval by Decoupling and Aggregation [22.803751188961865]
  We argue retrieval should move beyond similarity matching and instead operate over latent components.<n>We propose xMemory, which builds a hierarchy of intact units and maintains a searchable high-level node organisation.
  arXiv  Detail & Related papers  (2026-02-02T12:04:58Z)
• AMA: Adaptive Memory via Multi-Agent Collaboration [54.490349689939166]
  We propose Adaptive Memory via Multi-Agent Collaboration (AMA), a novel framework that leverages coordinated agents to manage memory across multiple granularities.<n>AMA significantly outperforms state-of-the-art baselines while reducing token consumption by approximately 80% compared to full-context methods.
  arXiv  Detail & Related papers  (2026-01-28T08:09:49Z)
• Agentic Learner with Grow-and-Refine Multimodal Semantic Memory [50.81667005063605]
  ViLoMem is a dual-stream memory framework that constructs compact, schema-based memory.<n>It encodes visual distraction patterns and logical reasoning errors, enabling MLLMs to learn from their successful and failed experiences.
  arXiv  Detail & Related papers  (2025-11-26T18:55:08Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.