arXiv:2608. 19564v1 Announce Type: new Abstract: Persistent memory can personalize an LLM agent, but an incorrect durable update can silently distort future behavior.
arXiv:2610.17763v1 Announce Type: new Abstract: Long-term memory supports agents across sessions. Repeated retrieval and write-back can let one historical claim occupy multiple retrieval positions an...
arXiv:2609.25054v1 Announce Type: new Abstract: For a long-horizon LLM agent, the memory question is not what was once recorded but what \emph{currently holds}. Most designs answer it only indirectly...
The paper investigates how memory systems can answer a current query correctly yet fail to retain distinctions needed for later updates. Using a paired‑history audit, the authors evaluate 24 history pairs across six synthetic mechanisms and two model backends, achieving perfect reveal accuracy on DeepSeek and high accuracy on GLM. Record‑level audits reveal specific failures in structured reveal memories and frontier late‑reference adequacy, and the authors test a label‑equivariant repair that only partially restores correctness.
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.
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.
The paper introduces two inference‑time designs—factor‑compiled admission (FC) and permission‑semantic admission (PS)—to govern persistent memory in large language model agents without retraining. FC evaluates entire memory entries, while PS decomposes entries into typed units, both using deterministic policies to decide what recalled information enters the working context and how it is expressed. Experiments on a development suite and an external benchmark show that FC and PS reduce failure rates and cross‑domain leakage compared to verbatim injection, though they increase personalization failures and do not meet all preregistered criteria.
The paper investigates how persistent memory in AI agents can lead to over‑trust in stale facts, creating a "Memory Trust Gap" that worsens as model capability increases. Using a benchmark with Benefit and Safety suites across Qwen3 models of varying sizes, the authors show that larger models are more prone to harmful over‑trust, especially when metadata is absent or misleading. They also demonstrate that mitigation strategies such as exposing metadata or pre‑resolving conflicts improve accuracy, but the effectiveness depends on model size and dataset.
The paper introduces the Distributed‑Evidence Paradox, where long‑running LLM agents compress past interactions into persistent memories that may not be fully supported by the interaction history. It defines three key requirements—evidence scope, compositional validity, and admission reliability—and proposes DerivAudit, a framework that checks whether a memory is truly supported by the available history. Experiments on two memory corpora show that expanding the evidence base can recover support for many memories, yet many remain unsupported, and broader evidence alone does not guarantee reliable admission.
The paper proposes a category‑conditioned retention strategy for agent memory, arguing that a single global confidence threshold is insufficient to balance reliability across different assertion types. By applying stricter retention bars to value assertions while allowing more liberal retention for other categories, the authors demonstrate a 36% relative reduction in unsupported value claims and a 13‑percentage‑point increase in coverage compared to a global threshold. The study is evaluated on a cold‑start memory pipeline with 100 synthetic personas, showing that selective, category‑aware retention improves overall memory reliability.
The paper introduces QEMem, a framework that rethinks how large language model agents use long‑term memory by constructing query‑specific evidence sets rather than relying on individual memory relevance. QEMem combines query‑guided discovery, source‑grounded reification into structured ViewCards, and sufficiency‑guided refinement to assemble comprehensive evidence from multiple memory granularities. Experiments show that QEMem achieves 91.43% LLM‑judge accuracy on LoCoMo with Claude Opus 4.5, surpassing leading baselines by significant margins and improving evidence coverage for cross‑session queries.
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.
The paper presents a retrieval‑admissibility verification framework for long‑term memory agents, classifying each memory‑query pair as admissible, inadmissible, or unresolved. It evaluates the framework on public benchmarks (RHELM and MemOps), showing improved anchor recall and reduced exact similarity errors, while also revealing that existing verifiers miss certain inadmissible exposures. The study highlights the need for separate checks on candidate support, admissibility, prompt exposure, and answer disclosure to ensure safe memory retrieval.