arXiv:2608. 19701v1 Announce Type: new Abstract: Long-term multi-agent systems continuously accumulate the memories produced by different agents.
Long-term multi-agent systems continuously accumulate the memories produced by different agents. Existing memory methods typically treat retrieved memories as independent evidence and combine them through voting or weighting.
arXiv:2610.33019v1 Announce Type: new Abstract: As multimodal agents engage in sustained interactions, long-term memory serves two distinct but complementary functions: fine-grained evidence recall a...
arXiv:2610.15642v1 Announce Type: new Abstract: Under a limited context budget, a long-term agent must recall not merely memories that are individually relevant, but evidence that is collectively suf...
Long-term memory has become increasingly important for LLM agents that operate across extended interactions and evolving task contexts. Recent memory systems have made past experiences more persistent, compact, and retrievable, but retrieval alone does not ensure that a memory provides valid evidence for the current query.
The paper investigates how large language model agents can over-rely on agentic memory, a phenomenon where retrieved memories distort inference even when they are correctly stored and retrieved. It shows that memory is helpful when past experience fully transfers to the current task but becomes misleading under partial query-memory overlap, a pattern confirmed by controlled experiments. To address this, the authors propose MEMTRIM, a plug‑and‑play framework that indexes memory evidence at write time and limits its reuse at read time, thereby reducing over-reliance without retraining and preserving useful memory benefits across models and memory architectures.
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.
MemFuseBench is a benchmark for multi‑source memory fusion that generates source‑tagged observations, evidence‑grounded questions, and adversarial distractors using a Scene‑to‑Sensor pipeline. The proposed MemFuse system preserves source‑level evidence in atomic memory and clusters related events into fused memory via a causal fusion graph, enabling traceable retrieval of dispersed observations. Experiments show that MemFuse outperforms other memory systems across all LLM settings, especially on questions requiring cross‑source evidence fusion.
LatentHarness unifies memory access and latent reasoning by treating them as sequential latent actions—THINK, RECALL, and EXIT—within a language model. It is trained via counterfactual policy distillation, which evaluates the impact of each action on the emitted token and learns when to recall evidence versus continue reasoning. On six long‑context reasoning benchmarks, a 1.4B‑parameter LatentHarness model outperforms the strongest baselines by 2.8% and 10.0% relative, while running 5.9× faster than the leading long‑context baseline.
arXiv:2606. 22030v2 Announce Type: replace Abstract: We investigate when belief-based memory actually improves large language model (LLM) agents.
MemoryLACE (MemLACE) is a lightweight memory framework that explicitly models the lifecycle of textual evidence—capturing sparse merge, supersession, and contradiction relations—while preserving atomic natural‑language memories and their provenance. Unlike traditional systems that retrieve memories independently, MemLACE reconstructs relation‑aware evidence units that expose current, historical, supporting, and conflicting evidence for downstream reasoning. In benchmark evaluations (BEAM and StructMemEval) using both open‑weight and proprietary LLM backbones, MemLACE achieves the highest overall performance among same‑backbone comparisons and reduces BEAM runtime by 66.6% compared to the strongest reflective‑memory baseline, Hindsight.
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.
arXiv:2608. 01285v1 Announce Type: new Abstract: The continued development of LLMs toward persistent and adaptive intelligence increasingly requires long-term memory mechanisms that preserve and reuse information across interactions.