Agent-ValueBench is the first dedicated benchmark for agent values, showing they diverge from LLM values, form a homogeneous 'Value Tide' across models, and bend under harnesses and skill steering.
super hub Canonical reference
Memory in the Age of AI Agents
Canonical reference. 82% of citing Pith papers cite this work as background.
104
Pith papers citing it
1
external citations · Pith
Background
82% of classified citations
abstract
Memory has emerged, and will continue to remain, a core capability of foundation model-based agents. As research on agent memory rapidly expands and attracts unprecedented attention, the field has also become increasingly fragmented. Existing works that fall under the umbrella of agent memory often differ substantially in their motivations, implementations, and evaluation protocols, while the proliferation of loosely defined memory terminologies has further obscured conceptual clarity. Traditional taxonomies such as long/short-term memory have proven insufficient to capture the diversity of contemporary agent memory systems. This work aims to provide an up-to-date landscape of current agent memory research. We begin by clearly delineating the scope of agent memory and distinguishing it from related concepts such as LLM memory, retrieval augmented generation (RAG), and context engineering. We then examine agent memory through the unified lenses of forms, functions, and dynamics. From the perspective of forms, we identify three dominant realizations of agent memory, namely token-level, parametric, and latent memory. From the perspective of functions, we propose a finer-grained taxonomy that distinguishes factual, experiential, and working memory. From the perspective of dynamics, we analyze how memory is formed, evolved, and retrieved over time. To support practical development, we compile a comprehensive summary of memory benchmarks and open-source frameworks. Beyond consolidation, we articulate a forward-looking perspective on emerging research frontiers, including memory automation, reinforcement learning integration, multimodal memory, multi-agent memory, and trustworthiness issues. We hope this survey serves not only as a reference for existing work, but also as a conceptual foundation for rethinking memory as a first-class primitive in the design of future agentic intelligence.
hub tools
citation-role summary
background 15
baseline 2
citation-polarity summary
claims ledger
- abstract Memory has emerged, and will continue to remain, a core capability of foundation model-based agents. As research on agent memory rapidly expands and attracts unprecedented attention, the field has also become increasingly fragmented. Existing works that fall under the umbrella of agent memory often differ substantially in their motivations, implementations, and evaluation protocols, while the proliferation of loosely defined memory terminologies has further obscured conceptual clarity. Traditional taxonomies such as long/short-term memory have proven insufficient to capture the diversity of co
authors
co-cited works
years
2026 104representative citing papers
A benchmark of 85 manually curated workplace scenarios reveals that multi-user AI agent systems suffer high rates of contextual integrity violations across outputs, inter-agent communication, and shared memory.
Introduces bounded-memory testbed for LLM agents in Slay the Spire 2 where typed retrieval replaces accumulating context, with released trajectories showing skill layer raises wins from 3/10 to 6/10.
MemSyco-Bench is a benchmark covering five tasks to evaluate memory-induced sycophancy in LLM agents, testing rejection of invalid memory, scope respect, conflict resolution, update tracking, and valid personalization.
MemTrace shows that evidence utilization, not retrieval, is the dominant failure mode in LLM long-term memory systems across tested configurations.
OSL-MR is a learning-augmented framework that casts memory retention as constrained stochastic optimization under partial observability and outperforms heuristic baselines on LoCoMo and LongMemEval.
M³Eval is a new cognitively-grounded benchmark that evaluates memory dimensions in multi-modal video models and reports consistent model weaknesses in disentanglement, interference, spatial-temporal grounding, and symbolic recall.
PersonaTree is a new hierarchical memory framework for persistent LLM agents that structures evidence into persona claims via support paths and outperforms baselines on six person-understanding benchmarks.
The study identifies four memory write channels and nine structural vulnerabilities in LLM agents, proposes a taxonomy of six attack classes, introduces MPBench, and finds that aggressive memory use increases exploitability while existing defenses fail.
eMEM is a multi-index memory architecture with tiered consolidation and ten recall tools for embodied agents, scoring 80.8 weighted mean on eMEM-Bench covering eight cognitive psychology paradigms and outperforming a flat RAG baseline on context and lure rejection tasks.
RHELM is a benchmark for LLM long-term memory with dynamic profiles, heterogeneous sources, and 27 memory characteristics that reveals weaknesses in existing models for multi-source aggregation and contextual reasoning.
MemPoison enables stealthy memory poisoning in LLM agents via dialogue by using semantic relational bridges, entity masquerading, and joint embedding optimization to bypass selective extraction and rewriting, achieving up to 0.95 attack success rate.
AgingBench demonstrates multi-dimensional degradation in deployed AI agents through four aging mechanisms diagnosed by temporal graphs and counterfactual probes across hundreds of runs.
Introduces PerMemBench benchmark for personalized memory and shows session-level gating yields retention gains under perfect decisions but accurate gating is an open challenge.
MemConflict provides a benchmark for testing LLM long-term memory systems under dynamic, static, and conditional conflicts involving temporal validity, factual correctness, and contextual applicability.
ClawForge is a generator framework that creates reproducible executable benchmarks for command-line agents under state conflict, with ClawForge-Bench showing frontier models reach at most 45.3% strict accuracy and that state inspection drives most performance gaps.
EvolveMem enables autonomous self-evolution of LLM memory retrieval configurations via LLM diagnosis and safeguards, delivering 25.7% gains over strong baselines on LoCoMo and 18.9% on MemBench with positive cross-benchmark transfer.
ScioMind combines anchoring-based belief updates, hierarchical memory, and dynamic profiles in LLM multi-agent systems to produce more stable, diverse, and psychologically aligned opinion trajectories than prior fixed-rule or unconstrained approaches.
Memory for long-horizon agents should preserve distinctions that affect decisions under a fixed budget, not descriptive features, yielding an exact forgetting boundary and a new online learner DeMem with regret guarantees.
A new evaluation protocol shows agent memory reliability degrades variably with added irrelevant sessions depending on agent, memory interface, and scale.
PropGuard is a propagation-aware framework for LLM-MAS that constructs dual-view spatio-temporal graphs, employs a GE-GRPO inspector to recover suspicious subgraphs, and applies source-guided remediation to lower attack success while preserving task performance.
SRTJ is a training-free jailbreak method that evolves hierarchical attack rules using iterative verifier feedback and ASP-based constraint-aware composition to achieve stable high success rates on HarmBench across multiple LLMs.
MemCoE learns memory organization guidelines via contrastive feedback and then trains a guideline-aligned RL policy for memory updates, yielding consistent gains on personalization benchmarks.
HeLa-Mem is a graph-based memory architecture for LLM agents that applies Hebbian learning to episodic associations and distills hubs into semantic knowledge, yielding better results on long-context benchmarks with fewer tokens.
citing papers explorer
-
Agent-ValueBench: A Comprehensive Benchmark for Evaluating Agent Values
Agent-ValueBench is the first dedicated benchmark for agent values, showing they diverge from LLM values, form a homogeneous 'Value Tide' across models, and bend under harnesses and skill steering.
-
PiSAs: Benchmarking Contextual Integrity in Multi-User Agentic Systems
A benchmark of 85 manually curated workplace scenarios reveals that multi-user AI agent systems suffer high rates of contextual integrity violations across outputs, inter-agent communication, and shared memory.
-
AgenticSTS: A Bounded-Memory Testbed for Long-Horizon LLM Agents
Introduces bounded-memory testbed for LLM agents in Slay the Spire 2 where typed retrieval replaces accumulating context, with released trajectories showing skill layer raises wins from 3/10 to 6/10.
-
MemSyco-Bench: Benchmarking Sycophancy in Agent Memory
MemSyco-Bench is a benchmark covering five tasks to evaluate memory-induced sycophancy in LLM agents, testing rejection of invalid memory, scope respect, conflict resolution, update tracking, and valid personalization.
-
MemTrace: Probing What Final Accuracy Misses in Long-Term Memory
MemTrace shows that evidence utilization, not retrieval, is the dominant failure mode in LLM long-term memory systems across tested configurations.
-
Learning What to Remember: Observability-Safe Memory Retention via Constrained Optimization for Long-Horizon Language Agents
OSL-MR is a learning-augmented framework that casts memory retention as constrained stochastic optimization under partial observability and outperforms heuristic baselines on LoCoMo and LongMemEval.
-
M$^3$Eval: Multi-Modal Memory Evaluation through Cognitively-Grounded Video Tasks
M³Eval is a new cognitively-grounded benchmark that evaluates memory dimensions in multi-modal video models and reports consistent model weaknesses in disentanglement, interference, spatial-temporal grounding, and symbolic recall.
-
PersonaTree: Structured Lifecycle Memory for Person Understanding in LLM Agents
PersonaTree is a new hierarchical memory framework for persistent LLM agents that structures evidence into persona claims via support paths and outperforms baselines on six person-understanding benchmarks.
-
From Untrusted Input to Trusted Memory: A Systematic Study of Memory Poisoning Attacks in LLM Agents
The study identifies four memory write channels and nine structural vulnerabilities in LLM agents, proposes a taxonomy of six attack classes, introduces MPBench, and finds that aggressive memory use increases exploitability while existing defenses fail.
-
eMEM: A Hybrid Spatio-Temporal Memory System For Embodied Agents
eMEM is a multi-index memory architecture with tiered consolidation and ten recall tools for embodied agents, scoring 80.8 weighted mean on eMEM-Bench covering eight cognitive psychology paradigms and outperforming a flat RAG baseline on context and lure rejection tasks.
-
Beyond Static Dialogues: Benchmarking Realistic, Heterogeneous, and Evolving Long-Term Memory
RHELM is a benchmark for LLM long-term memory with dynamic profiles, heterogeneous sources, and 27 memory characteristics that reveals weaknesses in existing models for multi-source aggregation and contextual reasoning.
-
Hijacking Agent Memory: Stealthy Trojan Attacks Through Conversational Interaction
MemPoison enables stealthy memory poisoning in LLM agents via dialogue by using semantic relational bridges, entity masquerading, and joint embedding optimization to bypass selective extraction and rewriting, achieving up to 0.95 attack success rate.
-
Your Agents Are Aging Too: Agent Lifespan Engineering for Deployed Systems
AgingBench demonstrates multi-dimensional degradation in deployed AI agents through four aging mechanisms diagnosed by temporal graphs and counterfactual probes across hundreds of runs.
-
Personalize-then-Store: Benchmarking and Learning Personalized Memory for Long-horizon Agents
Introduces PerMemBench benchmark for personalized memory and shows session-level gating yields retention gains under perfect decisions but accurate gating is an open challenge.
-
MemConflict: Evaluating Long-Term Memory Systems Under Memory Conflicts
MemConflict provides a benchmark for testing LLM long-term memory systems under dynamic, static, and conditional conflicts involving temporal validity, factual correctness, and contextual applicability.
-
ClawForge: Generating Executable Interactive Benchmarks for Command-Line Agents
ClawForge is a generator framework that creates reproducible executable benchmarks for command-line agents under state conflict, with ClawForge-Bench showing frontier models reach at most 45.3% strict accuracy and that state inspection drives most performance gaps.
-
EvolveMem:Self-Evolving Memory Architecture via AutoResearch for LLM Agents
EvolveMem enables autonomous self-evolution of LLM memory retrieval configurations via LLM diagnosis and safeguards, delivering 25.7% gains over strong baselines on LoCoMo and 18.9% on MemBench with positive cross-benchmark transfer.
-
ScioMind: Cognitively Grounded Multi-Agent Social Simulation with Anchoring-Based Belief Dynamics and Dynamic Profiles
ScioMind combines anchoring-based belief updates, hierarchical memory, and dynamic profiles in LLM multi-agent systems to produce more stable, diverse, and psychologically aligned opinion trajectories than prior fixed-rule or unconstrained approaches.
-
Remember the Decision, Not the Description: A Rate-Distortion Framework for Agent Memory
Memory for long-horizon agents should preserve distinctions that affect decisions under a fixed budget, not descriptive features, yielding an exact forgetting boundary and a new online learner DeMem with regret guarantees.
-
When Stored Evidence Stops Being Usable: Scale-Conditioned Evaluation of Agent Memory
A new evaluation protocol shows agent memory reliability degrades variably with added irrelevant sessions depending on agent, memory interface, and scale.
-
PropGuard: Safeguarding LLM-MAS via Propagation-Aware Exploration and Remediation
PropGuard is a propagation-aware framework for LLM-MAS that constructs dual-view spatio-temporal graphs, employs a GE-GRPO inspector to recover suspicious subgraphs, and applies source-guided remediation to lower attack success while preserving task performance.
-
SRTJ: Self-Evolving Rule-Driven Training-Free LLM Jailbreaking
SRTJ is a training-free jailbreak method that evolves hierarchical attack rules using iterative verifier feedback and ASP-based constraint-aware composition to achieve stable high success rates on HarmBench across multiple LLMs.
-
Learning How and What to Memorize: Cognition-Inspired Two-Stage Optimization for Evolving Memory
MemCoE learns memory organization guidelines via contrastive feedback and then trains a guideline-aligned RL policy for memory updates, yielding consistent gains on personalization benchmarks.
-
HeLa-Mem: Hebbian Learning and Associative Memory for LLM Agents
HeLa-Mem is a graph-based memory architecture for LLM agents that applies Hebbian learning to episodic associations and distills hubs into semantic knowledge, yielding better results on long-context benchmarks with fewer tokens.
-
When to Forget: A Memory Governance Primitive
Memory Worth converges almost surely to the conditional probability of task success given memory retrieval and correlates at rho=0.89 with ground-truth utility in controlled experiments.
-
PERMA: Benchmarking Personalized Memory Agents via Event-Driven Preference and Realistic Task Environments
PERMA is a new benchmark using temporally ordered events, text variability, and linguistic alignment to evaluate LLM memory agents on persona consistency beyond simple retrieval.
-
MemGround: Long-Term Memory Evaluation Kit for Large Language Models in Gamified Scenarios
MemGround is a new benchmark that evaluates LLMs' long-term memory through gamified tasks assessing surface state, temporal association, and reasoning memory.
-
CLAG: Adaptive Memory Organization via Agent-Driven Clustering for Small Language Model Agents
CLAG organizes agent memory into clusters via an SLM router and uses cluster profiles for two-stage retrieval, yielding better answer quality on QA benchmarks than prior memory systems.
-
What to Keep, What to Forget: A Rate--Distortion View of Memory Compaction in LLMs and Agents
KV-cache eviction, prompt compression, recurrent state bounding, and agent memory consolidation are unified as one rate-distortion problem with a shared lower bound, shared failure mode, and transferable mechanisms.
-
ACE: Pluggable Adaptive Context Elasticizer across Agents
ACE is a pluggable module that elastically orchestrates historical agent steps as raw, abstract, or dropped to maintain compact yet recoverable context for LLM agents handling long trajectories.
-
MemDelta: Controlled Baselines and Hidden Confounds in Agent Memory Evaluation
MemDelta shows agent memory evaluations are confounded by LLM family and embedding model, with RAG often matching full context and self-memory underperforming basic retrieval.
-
Neural Procedural Memory: Empowering LLM Agents with Implicit Activation Steering
NPM distills contrastive experiences into implicit activation steering vectors that guide LLM agent execution comparably to explicit RAG instructions, with complementary gains when combined.
-
Experience Graphs: The Data Foundation for Self-Improving Agents
Trellis treats agent experience graphs as first-class database state so that search patterns become queries, enabling crash recovery, scaling, and closed-loop training as architectural byproducts.
-
Mandol: An Agglomerative Agent Memory System for Long-Term Conversations
Mandol unifies memory storage and retrieval into an agglomerative semantic graph architecture with quantitative query mechanisms, reporting best accuracy on LoCoMo and LongMemEval plus 5.4x retrieval and 4.8x insertion speedups.
-
Analytic Concept-Centric Memory for Agentic Embodied Manipulation
Proposes a structured concept-centric memory system for embodied agents that connects object, scene, transition, and skill memories to support coarse-to-fine retrieval and improve task performance over baselines.
-
Empowering GUI Agents via Autonomous Experience Exploration and Hindsight Experience Utilization for Task Planning
PEEU enables a 7B MLLM to reach 30.6% accuracy on GUI task planning by autonomous exploration and hindsight experience synthesis, outperforming a 32B model through stronger high-level OOD generalization.
-
Are We Ready For An Agent-Native Memory System?
A four-module framework is used to benchmark 12 agent memory systems, showing no architecture dominates and that workload alignment plus localized maintenance drive performance and cost.
-
EvoEmbedding: Evolvable Representations for Long-Context Retrieval and Agentic Memory
EvoEmbedding generates evolvable embeddings via a latent memory updated during sequential processing, outperforming larger models on long-context retrieval and generalizing to 10x longer contexts in downstream tasks.
-
From Passive Generation to Investigation: A Proactive Scientific Peer Review Agent
ProReviewer is an MDP-formulated proactive peer review agent trained with SFT and RL on an 8B model that outperforms larger frontier LLMs on review quality metrics.
-
Toward Generalist Autonomous Research via Hypothesis-Tree Refinement
Arbor combines a coordinator, executors, and a hypothesis tree to enable cumulative autonomous research, outperforming Codex and Claude Code by over 2.5x on six real tasks and reaching 86.36% Any Medal on MLE-Bench Lite.
-
Beyond Similarity: Trustworthy Memory Search for Personal AI Agents
MemGate is a 9M-parameter neural gate inserted between vector memory and LLM that converts similarity search into task-conditioned admission, reducing memory-induced threats across agent frameworks while preserving utility.
-
Memory is Reconstructed, Not Retrieved: Graph Memory for LLM Agents
MRAgent combines a Cue-Tag-Content associative graph with active reconstruction to enable dynamic memory access in LLM agents, reporting up to 23% gains on long-memory benchmarks with lower token costs.
-
AdaMEM: Test-Time Adaptive Memory for Language Agents
AdaMEM proposes hybrid long-term and short-term memory for test-time adaptation in language agents, reporting relative gains of up to 13% on ALFWorld and 11% on WebShop over static baselines.
-
Harnessing Generalist Agents for Contextualized Time Series
TimeClaw is a framework that augments LLM agents with temporal tools, capability evolution, and episodic memory to enable contextualized time series reasoning, with reported gains on benchmarks across energy, finance, weather, and traffic.
-
Memory Shot for Long-Term Dialogue
MemShot renders local dialogue spans as structured visual memory units to improve long-term dialogue modeling in LLMs, achieving competitive benchmark performance with 70x faster memory construction.
-
Task-Focused Memorization for Multimodal Agents
TaskMem uses RL in two phases to learn a task-focused memorization policy for multimodal agents, yielding 5.3-7.0% VQA accuracy gains on reformulated streaming benchmarks from VideoMME, EgoLife, and EgoTempo.
-
Eywa: Provenance-Grounded Long-Term Memory for AI Agents
Eywa introduces a provenance-grounded memory system for persistent AI agents featuring evidence-first storage, typed validation, and deterministic multi-route retrieval, reporting 90.19% accuracy on LoCoMo and 88.2% on LongMemEval-S.
-
Rethinking Memory as Continuously Evolving Connectivity
FluxMem evolves memory as a heterogeneous graph via three refinement stages and reports consistent state-of-the-art results on LoCoMo, Mind2Web, and GAIA benchmarks.
-
The Labyrinth and the Thread: Rethinking Regularizations in Sequential Knowledge Editing for Large Language Models
Formal equivalence between one-time and sequential editing demonstrates that accumulated constraints suffice for stable LLM knowledge updates without specialized regularizations.
-
AgentFugue: Agent Scaling for Long-Horizon Tasks through Collective Reasoning
AgentFugue introduces a plug-in shared reasoning hub trained with SFT and RL that enables peer agents to share intermediate reasoning, yielding gains on long-horizon tasks over strong baselines.