Pith. sign in

REVIEW 2 minor 60 cited by

Agent memory research unifies under forms, functions, and dynamics with a new factual-experiential-working taxonomy.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

T0 review · grok-4.3

2026-05-11 18:10 UTC

load-bearing objection A useful survey that organizes agent memory work with a new three-lens taxonomy but introduces no new mechanisms or results.

arxiv 2512.13564 v2 submitted 2025-12-15 cs.CL cs.AI

Memory in the Age of AI Agents

classification cs.CL cs.AI
keywords AI agentsagent memoryfoundation modelsmemory taxonomyfactual memoryexperiential memoryworking memorymemory dynamics
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

AI agents built on foundation models depend on memory systems that have proliferated rapidly yet remain conceptually scattered across motivations, implementations, and terms. This survey first separates agent memory from nearby ideas such as LLM memory and retrieval-augmented generation. It then applies three consistent lenses—forms, functions, and dynamics—to organize the literature. From the function lens it introduces a finer taxonomy that replaces broad long-term versus short-term labels with factual, experiential, and working memory categories. The result supplies a consolidated reference, a list of benchmarks and open frameworks, and a map of open research directions.

Core claim

This survey delineates the scope of agent memory and examines it through the unified lenses of forms (token-level, parametric, and latent realizations), functions (factual, experiential, and working memory), and dynamics (how memory is formed, evolved, and retrieved). It argues that traditional long/short-term distinctions are insufficient for contemporary agent systems and compiles benchmarks, frameworks, and forward-looking topics such as memory automation, reinforcement-learning integration, multimodal memory, multi-agent memory, and trustworthiness to support memory as a first-class design primitive.

What carries the argument

The three lenses of forms, functions, and dynamics, with the function-based taxonomy that distinguishes factual, experiential, and working memory.

Load-bearing premise

The distinctions among forms, functions, and dynamics form a complete, non-overlapping classification that meaningfully reduces fragmentation in the existing literature.

What would settle it

A later systematic mapping of published agent systems that shows most implementations still fall outside the factual-experiential-working categories or require substantial overlap would falsify the taxonomy's claimed unifying power.

Watch this falsifier — get emailed when new claim-graph text bears on it.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

0 major / 2 minor

Summary. The paper is a survey on memory systems for foundation model-based AI agents. It argues that the field is fragmented with proliferating terminologies and that traditional long/short-term memory distinctions are insufficient. The authors delineate the scope of agent memory from related concepts such as LLM memory, RAG, and context engineering; propose taxonomies organized by forms (token-level, parametric, latent), functions (factual, experiential, working), and dynamics (formation, evolution, retrieval); compile benchmarks and open-source frameworks; and outline future frontiers including memory automation, RL integration, multimodal memory, multi-agent memory, and trustworthiness issues.

Significance. If the taxonomy is adopted, the survey could meaningfully consolidate a rapidly expanding area by supplying a unified organizational lens that better captures contemporary agent memory systems than prior distinctions. The explicit compilation of benchmarks and frameworks provides immediate practical utility for researchers and developers, while the forward-looking section on emerging frontiers offers a useful roadmap. These elements position the work as a potential reference point for treating memory as a first-class design primitive in agentic systems.

minor comments (2)
  1. [Abstract] The abstract states that the survey compiles 'a comprehensive summary of memory benchmarks and open-source frameworks' but does not indicate selection criteria or coverage scope; adding a short methods paragraph or table in the main text would improve reproducibility and transparency of the consolidation effort.
  2. [Scope delineation] The scope delineation from RAG and context engineering is conceptually useful; a concise comparative table (e.g., in the introduction) listing key differences in motivation, implementation, and evaluation would enhance clarity without altering the central argument.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive and constructive review, which highlights the potential of the proposed taxonomy, benchmark compilation, and future directions to consolidate the agent memory literature. We appreciate the recommendation for minor revision.

Circularity Check

0 steps flagged

No significant circularity; survey taxonomy is externally grounded

full rationale

This is a survey paper whose central contribution is an organizational taxonomy of agent memory drawn from analysis of external literature. It delineates scope against related concepts (LLM memory, RAG, context engineering), identifies forms (token-level/parametric/latent), proposes functions (factual/experiential/working), and examines dynamics without any equations, fitted parameters, predictions, or derivations. No load-bearing step reduces to self-definition, self-citation chains, or renaming of known results by construction; the distinctions are explicitly motivated as a response to fragmentation in prior work. The paper is self-contained against external benchmarks and compiles summaries of existing frameworks rather than deriving new results from its own inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

This is a survey that consolidates and reorganizes existing research rather than deriving new results from first principles or introducing new entities. It relies on the domain assumption that current literature is sufficiently fragmented to require a new taxonomy.

axioms (2)
  • domain assumption Traditional taxonomies such as long/short-term memory are insufficient to capture the diversity of contemporary agent memory systems
    Explicitly stated in the abstract as motivation for the new framework.
  • domain assumption Agent memory is distinct from LLM memory, RAG, and context engineering
    Stated as the starting point for delineating scope.

pith-pipeline@v0.9.0 · 5769 in / 1437 out tokens · 115660 ms · 2026-05-11T18:10:58.826346+00:00 · methodology

0 comments
read the original 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.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 60 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Agent-ValueBench: A Comprehensive Benchmark for Evaluating Agent Values

    cs.AI 2026-05 unverdicted novelty 8.0

    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.

  2. Trojan Hippo: Weaponizing Agent Memory for Data Exfiltration

    cs.CR 2026-05 unverdicted novelty 8.0

    Trojan Hippo attacks on LLM agent memory achieve 85-100% success rates in data exfiltration across four memory backends even after 100 benign sessions, while evaluated defenses reduce success rates but impose varying ...

  3. AgenticSTS: A Bounded-Memory Testbed for Long-Horizon LLM Agents

    cs.AI 2026-07 unverdicted novelty 7.0

    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.

  4. MemSyco-Bench: Benchmarking Sycophancy in Agent Memory

    cs.IR 2026-07 unverdicted novelty 7.0

    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.

  5. MemSyco-Bench: Benchmarking Sycophancy in Agent Memory

    cs.IR 2026-07 unverdicted novelty 7.0

    MemSyco-Bench is a new benchmark with five tasks to assess memory-induced sycophancy in LLM agent systems.

  6. MemTrace: Probing What Final Accuracy Misses in Long-Term Memory

    cs.AI 2026-06 unverdicted novelty 7.0

    MemTrace shows that evidence utilization, not retrieval, is the dominant failure mode in LLM long-term memory systems across tested configurations.

  7. Learning What to Remember: Observability-Safe Memory Retention via Constrained Optimization for Long-Horizon Language Agents

    cs.AI 2026-06 unverdicted novelty 7.0

    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.

  8. M$^3$Eval: Multi-Modal Memory Evaluation through Cognitively-Grounded Video Tasks

    cs.CV 2026-06 unverdicted novelty 7.0

    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 sym...

  9. PersonaTree: Structured Lifecycle Memory for Person Understanding in LLM Agents

    cs.CL 2026-06 unverdicted novelty 7.0

    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.

  10. From Untrusted Input to Trusted Memory: A Systematic Study of Memory Poisoning Attacks in LLM Agents

    cs.CR 2026-06 unverdicted novelty 7.0

    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 exploita...

  11. eMEM: A Hybrid Spatio-Temporal Memory System For Embodied Agents

    cs.RO 2026-06 unverdicted novelty 7.0

    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 ...

  12. Beyond Static Dialogues: Benchmarking Realistic, Heterogeneous, and Evolving Long-Term Memory

    cs.CL 2026-05 unverdicted novelty 7.0

    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.

  13. Hijacking Agent Memory: Stealthy Trojan Attacks Through Conversational Interaction

    cs.CR 2026-05 unverdicted novelty 7.0

    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, achievin...

  14. Your Agents Are Aging Too: Agent Lifespan Engineering for Deployed Systems

    cs.AI 2026-05 unverdicted novelty 7.0

    AgingBench demonstrates multi-dimensional degradation in deployed AI agents through four aging mechanisms diagnosed by temporal graphs and counterfactual probes across hundreds of runs.

  15. Personalize-then-Store: Benchmarking and Learning Personalized Memory for Long-horizon Agents

    cs.AI 2026-05 unverdicted novelty 7.0

    Introduces PerMemBench benchmark for personalized memory and shows session-level gating yields retention gains under perfect decisions but accurate gating is an open challenge.

  16. MemConflict: Evaluating Long-Term Memory Systems Under Memory Conflicts

    cs.IR 2026-05 unverdicted novelty 7.0

    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.

  17. ClawForge: Generating Executable Interactive Benchmarks for Command-Line Agents

    cs.AI 2026-05 unverdicted novelty 7.0

    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 tha...

  18. ClawForge: Generating Executable Interactive Benchmarks for Command-Line Agents

    cs.AI 2026-05 conditional novelty 7.0

    ClawForge supplies a generator that turns scenario templates into reproducible command-line tasks testing state conflict handling, where the strongest frontier model scores only 45.3 percent strict accuracy.

  19. EvolveMem:Self-Evolving Memory Architecture via AutoResearch for LLM Agents

    cs.LG 2026-05 unverdicted novelty 7.0

    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-ben...

  20. ScioMind: Cognitively Grounded Multi-Agent Social Simulation with Anchoring-Based Belief Dynamics and Dynamic Profiles

    cs.AI 2026-05 unverdicted novelty 7.0

    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...

  21. Remember the Decision, Not the Description: A Rate-Distortion Framework for Agent Memory

    cs.AI 2026-05 unverdicted novelty 7.0

    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.

  22. When Stored Evidence Stops Being Usable: Scale-Conditioned Evaluation of Agent Memory

    cs.AI 2026-05 unverdicted novelty 7.0

    A new evaluation protocol shows agent memory reliability degrades variably with added irrelevant sessions depending on agent, memory interface, and scale.

  23. PropGuard: Safeguarding LLM-MAS via Propagation-Aware Exploration and Remediation

    cs.LG 2026-05 unverdicted novelty 7.0

    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 atta...

  24. SRTJ: Self-Evolving Rule-Driven Training-Free LLM Jailbreaking

    cs.CR 2026-05 unverdicted novelty 7.0

    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 mul...

  25. Learning How and What to Memorize: Cognition-Inspired Two-Stage Optimization for Evolving Memory

    cs.CL 2026-05 unverdicted novelty 7.0

    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.

  26. Order-invariant cluster first-order logic on graph classes of bounded degree

    cs.LO 2026-04 unverdicted novelty 7.0

    Order-invariant cluster first-order logic has the same expressive power as first-order logic on every class of graphs with bounded degree.

  27. HeLa-Mem: Hebbian Learning and Associative Memory for LLM Agents

    cs.CL 2026-04 unverdicted novelty 7.0

    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 fe...

  28. When to Forget: A Memory Governance Primitive

    cs.AI 2026-04 unverdicted novelty 7.0

    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.

  29. PERMA: Benchmarking Personalized Memory Agents via Event-Driven Preference and Realistic Task Environments

    cs.AI 2026-03 unverdicted novelty 7.0

    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.

  30. MemGround: Long-Term Memory Evaluation Kit for Large Language Models in Gamified Scenarios

    cs.CL 2026-03 unverdicted novelty 7.0

    MemGround is a new benchmark that evaluates LLMs' long-term memory through gamified tasks assessing surface state, temporal association, and reasoning memory.

  31. CLAG: Adaptive Memory Organization via Agent-Driven Clustering for Small Language Model Agents

    cs.CL 2026-03 unverdicted novelty 7.0

    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.

  32. ACE: Pluggable Adaptive Context Elasticizer across Agents

    cs.AI 2026-06 unverdicted novelty 6.0

    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.

  33. MemDelta: Controlled Baselines and Hidden Confounds in Agent Memory Evaluation

    cs.CL 2026-06 unverdicted novelty 6.0

    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.

  34. Neural Procedural Memory: Empowering LLM Agents with Implicit Activation Steering

    cs.CL 2026-06 unverdicted novelty 6.0

    NPM distills contrastive experiences into implicit activation steering vectors that guide LLM agent execution comparably to explicit RAG instructions, with complementary gains when combined.

  35. Experience Graphs: The Data Foundation for Self-Improving Agents

    cs.DB 2026-06 unverdicted novelty 6.0

    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.

  36. Mandol: An Agglomerative Agent Memory System for Long-Term Conversations

    cs.DB 2026-06 unverdicted novelty 6.0

    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 insertio...

  37. Analytic Concept-Centric Memory for Agentic Embodied Manipulation

    cs.RO 2026-06 unverdicted novelty 6.0

    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.

  38. Empowering GUI Agents via Autonomous Experience Exploration and Hindsight Experience Utilization for Task Planning

    cs.CL 2026-06 unverdicted novelty 6.0

    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.

  39. Are We Ready For An Agent-Native Memory System?

    cs.CL 2026-06 unverdicted novelty 6.0

    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.

  40. EvoEmbedding: Evolvable Representations for Long-Context Retrieval and Agentic Memory

    cs.CL 2026-06 unverdicted novelty 6.0

    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.

  41. From Passive Generation to Investigation: A Proactive Scientific Peer Review Agent

    cs.CL 2026-06 unverdicted novelty 6.0

    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.

  42. Toward Generalist Autonomous Research via Hypothesis-Tree Refinement

    cs.CL 2026-06 unverdicted novelty 6.0

    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.

  43. Learning What to Remember: Observability-Safe Memory Retention via Constrained Optimization for Long-Horizon Language Agents

    cs.AI 2026-06 unverdicted novelty 6.0

    OSL-MR applies constrained stochastic optimization and learning to memory retention in long-horizon agents, outperforming recency and heuristic baselines on LoCoMo and LongMemEval under tight budgets.

  44. Beyond Similarity: Trustworthy Memory Search for Personal AI Agents

    cs.AI 2026-06 unverdicted novelty 6.0

    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.

  45. Memory is Reconstructed, Not Retrieved: Graph Memory for LLM Agents

    cs.AI 2026-06 unverdicted novelty 6.0

    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.

  46. AdaMEM: Test-Time Adaptive Memory for Language Agents

    cs.AI 2026-06 unverdicted novelty 6.0

    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.

  47. Harnessing Generalist Agents for Contextualized Time Series

    cs.AI 2026-06 unverdicted novelty 6.0

    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,...

  48. Memory Shot for Long-Term Dialogue

    cs.IR 2026-05 unverdicted novelty 6.0

    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.

  49. Task-Focused Memorization for Multimodal Agents

    cs.CV 2026-05 unverdicted novelty 6.0

    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.

  50. Eywa: Provenance-Grounded Long-Term Memory for AI Agents

    cs.CL 2026-05 unverdicted novelty 6.0

    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% o...

  51. Rethinking Memory as Continuously Evolving Connectivity

    cs.CL 2026-05 unverdicted novelty 6.0

    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.

  52. The Labyrinth and the Thread: Rethinking Regularizations in Sequential Knowledge Editing for Large Language Models

    cs.CL 2026-05 unverdicted novelty 6.0

    Formal equivalence between one-time and sequential editing demonstrates that accumulated constraints suffice for stable LLM knowledge updates without specialized regularizations.

  53. AgentFugue: Agent Scaling for Long-Horizon Tasks through Collective Reasoning

    cs.AI 2026-05 unverdicted novelty 6.0

    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.

  54. SAM: State-Adaptive Memory for Long-Horizon Reasoning Agent

    cs.AI 2026-05 unverdicted novelty 6.0

    SAM is a standalone memory framework for long-horizon LLM agents that creates state-adaptive cues from interactions, preserves raw trajectories for intent-driven recall, and optimizes the module via expert supervision...

  55. DeliCIR: Deliberative Test-Time Evolutionary Hierarchical Multi-Agents for Composed Image Retrieval

    cs.CV 2026-05 unverdicted novelty 6.0

    Proposes PDF, a hierarchical multi-agent Perception-to-Deliberation Framework that adds experience self-evolution and test-time scaling to composed image retrieval, claiming SOTA on CIRR, CIRCO, and FashionIQ.

  56. DeliCIR: Deliberative Test-Time Evolutionary Hierarchical Multi-Agents for Composed Image Retrieval

    cs.CV 2026-05 unverdicted novelty 6.0

    PDF introduces a hierarchical Perception-to-Deliberation Framework with multi-agents, intent routing, and tournament-style test-time scaling to achieve SOTA results on ZS-CIR benchmarks CIRR, CIRCO, and FashionIQ.

  57. DeferMem: Query-Time Evidence Distillation via Reinforcement Learning for Long-Term Memory QA

    cs.CL 2026-05 unverdicted novelty 6.0

    DeferMem decouples memory QA into high-recall retrieval and RL-based query-conditioned evidence distillation, outperforming baselines on LoCoMo and LongMemEval-S with highest accuracy, fastest runtime, and zero API to...

  58. Mem-$\pi$: Adaptive Memory through Learning When and What to Generate

    cs.CL 2026-05 unverdicted novelty 6.0

    Mem-π is a framework using a dedicated model and decision-content decoupled RL to generate context-specific guidance on demand for LLM agents, outperforming retrieval baselines by over 30% on web navigation.

  59. Auto-Dreamer: Learning Offline Memory Consolidation for Language Agents

    cs.CL 2026-05 unverdicted novelty 6.0

    Auto-Dreamer trains an offline memory consolidator via GRPO on agent performance to abstract cross-session patterns, outperforming baselines by 7 points on ScienceWorld with 12x smaller memory and generalizing to ALFW...

  60. Remembering More, Risking More: Longitudinal Safety Risks in Memory-Equipped LLM Agents

    cs.AI 2026-05 unverdicted novelty 6.0

    Memory-equipped LLM agents exhibit increasing safety violation rates as memory accumulates across independent tasks, termed temporal memory contamination, detected via a new trigger-probe protocol.

Reference graph

Works this paper leans on

300 extracted references · 300 canonical work pages · cited by 102 Pith papers · 16 internal anchors

  1. [1]

    Detoxifying Large Language Models via Knowledge Editing

    Mengru Wang and Ningyu Zhang and Ziwen Xu and Zekun Xi and Shumin Deng and Yunzhi Yao and Qishen Zhang and Linyi Yang and Jindong Wang and Huajun Chen , editor =. Detoxifying Large Language Models via Knowledge Editing , booktitle =. 2024 , url =. doi:10.18653/V1/2024.ACL-LONG.171 , timestamp =

  2. [2]

    Neighboring Perturbations of Knowledge Editing on Large Language Models , booktitle =

    Jun. Neighboring Perturbations of Knowledge Editing on Large Language Models , booktitle =. 2024 , url =

  3. [3]

    Editing Personality For Large Language Models , booktitle =

    Shengyu Mao and Xiaohan Wang and Mengru Wang and Yong Jiang and Pengjun Xie and Fei Huang and Ningyu Zhang , editor =. Editing Personality For Large Language Models , booktitle =. 2024 , url =. doi:10.1007/978-981-97-9434-8\_19 , timestamp =

  4. [4]

    Manning , title =

    Eric Mitchell and Charles Lin and Antoine Bosselut and Chelsea Finn and Christopher D. Manning , title =. The Tenth International Conference on Learning Representations,. 2022 , url =

  5. [5]

    The Twelfth International Conference on Learning Representations,

    Guangxuan Xiao and Yuandong Tian and Beidi Chen and Song Han and Mike Lewis , title =. The Twelfth International Conference on Learning Representations,. 2024 , url =

  6. [6]

    Zep: A Temporal Knowledge Graph Architecture for Agent Memory

    Preston Rasmussen and Pavlo Paliychuk and Travis Beauvais and Jack Ryan and Daniel Chalef , title =. CoRR , volume =. 2025 , url =. doi:10.48550/ARXIV.2501.13956 , eprinttype =. 2501.13956 , timestamp =

  7. [7]

    arXiv preprint arXiv:2401.04658 , year =

    Zhen Qin and Weigao Sun and Dong Li and Xuyang Shen and Weixuan Sun and Yiran Zhong , title =. CoRR , volume =. 2024 , url =. doi:10.48550/ARXIV.2401.04658 , eprinttype =. 2401.04658 , timestamp =

  8. [8]

    Forty-first International Conference on Machine Learning,

    Zhen Qin and Weigao Sun and Dong Li and Xuyang Shen and Weixuan Sun and Yiran Zhong , title =. Forty-first International Conference on Machine Learning,. 2024 , url =

  9. [9]

    FlashAttention-3: Fast and Accurate Attention with Asynchrony and Low-precision , booktitle =

    Jay Shah and Ganesh Bikshandi and Ying Zhang and Vijay Thakkar and Pradeep Ramani and Tri Dao , editor =. FlashAttention-3: Fast and Accurate Attention with Asynchrony and Low-precision , booktitle =. 2024 , url =

  10. [10]

    The Twelfth International Conference on Learning Representations,

    Tri Dao , title =. The Twelfth International Conference on Learning Representations,. 2024 , url =

  11. [11]

    Rethinking Attention with Performers , booktitle =

    Krzysztof Marcin Choromanski and Valerii Likhosherstov and David Dohan and Xingyou Song and Andreea Gane and Tam. Rethinking Attention with Performers , booktitle =. 2021 , url =

  12. [12]

    Big Bird: Transformers for Longer Sequences , booktitle =

    Manzil Zaheer and Guru Guruganesh and Kumar Avinava Dubey and Joshua Ainslie and Chris Alberti and Santiago Onta. Big Bird: Transformers for Longer Sequences , booktitle =. 2020 , url =

  13. [13]

    In: IEEE Winter Conf

    Qiuhui Chen and Qiang Fu and Hao Bai and Yi Hong , title =. 2024 , url =. doi:10.1109/WACV57701.2024.00354 , timestamp =

  14. [14]

    A Machine with Short-Term, Episodic, and Semantic Memory Systems , booktitle =

    Taewoon Kim and Michael Cochez and Vincent Fran. A Machine with Short-Term, Episodic, and Semantic Memory Systems , booktitle =. 2023 , url =. doi:10.1609/AAAI.V37I1.25075 , timestamp =

  15. [15]

    McAuley , title =

    Yu Wang and Xinshuang Liu and Xiusi Chen and Sean O'Brien and Junda Wu and Julian J. McAuley , title =. The Thirteenth International Conference on Learning Representations,. 2025 , url =

  16. [16]

    Agents: An Open-source Framework for Autonomous Language Agents, 2023

    Wangchunshu Zhou and Yuchen Eleanor Jiang and Long Li and Jialong Wu and Tiannan Wang and Shi Qiu and Jintian Zhang and Jing Chen and Ruipu Wu and Shuai Wang and Shiding Zhu and Jiyu Chen and Wentao Zhang and Ningyu Zhang and Huajun Chen and Peng Cui and Mrinmaya Sachan , title =. CoRR , volume =. 2023 , url =. doi:10.48550/ARXIV.2309.07870 , eprinttype =...

  17. [17]

    OA gents: An Empirical Study of Building Effective Agents

    Zhu, He and Qin, Tianrui and Zhu, King and Huang, Heyuan and Guan, Yeyi and Xia, Jinxiang and Li, Hanhao and Yao, Yi and Wang, Ningning and Liu, Pai and Peng, Tianhao and Gui, Xin and Xiaowan, Li and Liu, Yuhui and Tang, Xiangru and Yang, Jian and Zhang, Ge and Gao, Xitong and Jiang, Yuchen Eleanor and Zhang, Changwang and Wang, Jun and Liu, Jiaheng and Z...

  18. [18]

    2025 , eprint=

    Flash-Searcher: Fast and Effective Web Agents via DAG-Based Parallel Execution , author=. 2025 , eprint=

  19. [19]

    2025 , eprint=

    Towards Personalized Deep Research: Benchmarks and Evaluations , author=. 2025 , eprint=

  20. [20]

    2025 , eprint=

    Chain-of-Agents: End-to-End Agent Foundation Models via Multi-Agent Distillation and Agentic RL , author=. 2025 , eprint=

  21. [21]

    2023 , eprint=

    RecurrentGPT: Interactive Generation of (Arbitrarily) Long Text , author=. 2023 , eprint=

  22. [22]

    2024 , eprint=

    AI PERSONA: Towards Life-long Personalization of LLMs , author=. 2024 , eprint=

  23. [23]

    arXiv preprint arXiv:2406.18532 , year=

    Wangchunshu Zhou and Yixin Ou and Shengwei Ding and Long Li and Jialong Wu and Tiannan Wang and Jiamin Chen and Shuai Wang and Xiaohua Xu and Ningyu Zhang and Huajun Chen and Yuchen Eleanor Jiang , title =. CoRR , volume =. 2024 , url =. doi:10.48550/ARXIV.2406.18532 , eprinttype =. 2406.18532 , timestamp =

  24. [24]

    2025 , eprint=

    EvoVLA: Self-Evolving Vision-Language-Action Model , author=. 2025 , eprint=

  25. [25]

    2025 , eprint=

    O-Mem: Omni Memory System for Personalized, Long Horizon, Self-Evolving Agents , author=. 2025 , eprint=

  26. [26]

    MLP memory: A retriever-pretrained memory for large language models.arXiv preprint arXiv:2508.01832, 2025

    Rubin Wei and Jiaqi Cao and Jiarui Wang and Jushi Kai and Qipeng Guo and Bowen Zhou and Zhouhan Lin , title =. CoRR , volume =. 2025 , url =. doi:10.48550/ARXIV.2508.01832 , eprinttype =. 2508.01832 , timestamp =

  27. [27]

    2025 , eprint=

    Pretraining with hierarchical memories: separating long-tail and common knowledge , author=. 2025 , eprint=

  28. [28]

    Character-LLM:

    Yunfan Shao and Linyang Li and Junqi Dai and Xipeng Qiu , editor =. Character-LLM:. Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing,. 2023 , url =. doi:10.18653/V1/2023.EMNLP-MAIN.814 , timestamp =

  29. [29]

    High”, skill “Proficient

    Jinfeng Zhou and Zhuang Chen and Dazhen Wan and Bosi Wen and Yi Song and Jifan Yu and Yongkang Huang and Pei Ke and Guanqun Bi and Libiao Peng and Jiaming Yang and Xiyao Xiao and Sahand Sabour and Xiaohan Zhang and Wenjing Hou and Yijia Zhang and Yuxiao Dong and Hongning Wang and Jie Tang and Minlie Huang , editor =. CharacterGLM: Customizing Social Chara...

  30. [30]

    2025 , eprint=

    Agent Learning via Early Experience , author=. 2025 , eprint=

  31. [31]

    Scaling agents via continual pre-training.arXiv preprint arXiv:2509.13310,

    Liangcai Su and Zhen Zhang and Guangyu Li and Zhuo Chen and Chenxi Wang and Maojia Song and Xinyu Wang and Kuan Li and Jialong Wu and Xuanzhong Chen and Zile Qiao and Zhongwang Zhang and Huifeng Yin and Shihao Cai and Runnan Fang and Zhengwei Tao and Wenbiao Yin and Chenxiong Qian and Yong Jiang and Pengjun Xie and Fei Huang and Jingren Zhou , title =. Co...

  32. [32]

    Online Adaptation of Language Models with a Memory of Amortized Contexts , booktitle =

    Jihoon Tack and Jaehyung Kim and Eric Mitchell and Jinwoo Shin and Yee Whye Teh and Jonathan Richard Schwarz , editor =. Online Adaptation of Language Models with a Memory of Amortized Contexts , booktitle =. 2024 , url =

  33. [33]

    Advances and Challenges in Foundation Agents: From Brain-Inspired Intelligence to Evolutionary, Collaborative, and Safe Systems

    Advances and challenges in foundation agents: From brain-inspired intelligence to evolutionary, collaborative, and safe systems , author=. arXiv preprint arXiv:2504.01990 , year=

  34. [34]

    2025 , eprint=

    From AI for Science to Agentic Science: A Survey on Autonomous Scientific Discovery , author=. 2025 , eprint=

  35. [35]

    Tool learning with large language models: a survey

    Qu, Changle and Dai, Sunhao and Wei, Xiaochi and Cai, Hengyi and Wang, Shuaiqiang and Yin, Dawei and Xu, Jun and Wen, Ji-rong , year=. Tool learning with large language models: a survey , volume=. Frontiers of Computer Science , publisher=. doi:10.1007/s11704-024-40678-2 , number=

  36. [36]

    2025 , eprint=

    Reinforcement Learning for Reasoning in Large Language Models with One Training Example , author=. 2025 , eprint=

  37. [37]

    2025 , eprint=

    A Comprehensive Survey of Self-Evolving AI Agents: A New Paradigm Bridging Foundation Models and Lifelong Agentic Systems , author=. 2025 , eprint=

  38. [38]

    2024 , eprint=

    Agent AI: Surveying the Horizons of Multimodal Interaction , author=. 2024 , eprint=

  39. [39]

    2024 , eprint=

    Agents in Software Engineering: Survey, Landscape, and Vision , author=. 2024 , eprint=

  40. [40]

    2025 , eprint=

    Deep Research: A Survey of Autonomous Research Agents , author=. 2025 , eprint=

  41. [41]

    2025 , eprint=

    A Comprehensive Survey of Deep Research: Systems, Methodologies, and Applications , author=. 2025 , eprint=

  42. [42]

    2025 , eprint=

    Large Language Model Agent: A Survey on Methodology, Applications and Challenges , author=. 2025 , eprint=

  43. [43]

    Zhao, Andrew and Wu, Yiran and Yue, Yang and Wu, Tong and Xu, Quentin and Lin, Matthieu and Wang, Shenzhi and Wu, Qingyun and Zheng, Zilong and Huang, Gao , journal=

  44. [44]

    2025 , eprint=

    Large Language Models: A Survey , author=. 2025 , eprint=

  45. [45]

    2025 , eprint=

    A Survey on Large Language Models with some Insights on their Capabilities and Limitations , author=. 2025 , eprint=

  46. [46]

    Huang, Chengsong and Yu, Wenhao and Wang, Xiaoyang and Zhang, Hongming and Li, Zongxia and Li, Ruosen and Huang, Jiaxin and Mi, Haitao and Yu, Dong , journal=

  47. [47]

    DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models

    Deepseekmath: Pushing the limits of mathematical reasoning in open language models , author=. arXiv preprint arXiv:2402.03300 , year=

  48. [48]

    First Conference on Language Modeling , year=

    Autogen: Enabling next-gen LLM applications via multi-agent conversations , author=. First Conference on Language Modeling , year=

  49. [49]

    Wang, Yingxu and Liu, Siwei and Fang, Jinyuan and Meng, Zaiqiao , journal=

  50. [50]

    Advances in Neural Information Processing Systems , year =

    Shuofei Qiao and Runnan Fang and Ningyu Zhang and Yuqi Zhu and Xiang Chen and Shumin Deng and Yong Jiang and Pengjun Xie and Fei Huang and Huajun Chen , title =. Advances in Neural Information Processing Systems , year =

  51. [51]

    arXiv preprint arXiv:2507.21407

    Graph-Augmented Large Language Model Agents: Current Progress and Future Prospects , author=. arXiv preprint arXiv:2507.21407 , year=

  52. [52]

    Yu, Miao and Wang, Shilong and Zhang, Guibin and Mao, Junyuan and Yin, Chenlong and Liu, Qijiong and Wen, Qingsong and Wang, Kun and Wang, Yang , journal=

  53. [53]

    Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages =

    Yanwei Yue and Guibin Zhang and Boyang Liu and Guancheng Wan and Kun Wang and Dawei Cheng and Yiyan Qi , title =. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages =

  54. [54]

    Zhang, Guibin and Fu, Muxin and Wan, Guancheng and Yu, Miao and Wang, Kun and Yan, Shuicheng , journal=

  55. [55]

    Liu, Siwei and Fang, Jinyuan and Zhou, Han and Wang, Yingxu and Meng, Zaiqiao , journal=

  56. [56]

    ACM Transactions on Modeling and Computer Simulation 35 (4), 1-17

    Reasoning RAG via System 1 or System 2: A Survey on Reasoning Agentic Retrieval-Augmented Generation for Industry Challenges , author=. arXiv preprint arXiv:2506.10408 , year=

  57. [57]

    Liao, Junwei and Wen, Muning and Wang, Jun and Zhang, Weinan , journal=

  58. [58]

    Ozdaglar and Kaiqing Zhang and Joo

    Chanwoo Park and Seungju Han and Xingzhi Guo and Asuman E. Ozdaglar and Kaiqing Zhang and Joo. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages =

  59. [59]

    Zhao, Wanjia and Yuksekgonul, Mert and Wu, Shirley and Zou, James , journal=

  60. [60]

    Survey of

    Sarkar, Anjana and Sarkar, Soumyendu , journal=. Survey of

  61. [61]

    CoRR , year=

    Enhancing the Code Debugging Ability of LLMs via Communicative Agent Based Data Refinement , author=. CoRR , year=

  62. [62]

    Yang, Yingxuan and Chai, Huacan and Shao, Shuai and Song, Yuanyi and Qi, Siyuan and Rui, Renting and Zhang, Weinan , journal=

  63. [63]

    A survey of

    Yang, Yingxuan and Chai, Huacan and Song, Yuanyi and Qi, Siyuan and Wen, Muning and Li, Ning and Liao, Junwei and Hu, Haoyi and Lin, Jianghao and Chang, Gaowei and others , journal=. A survey of

  64. [64]

    Yao, Shunyu and Zhao, Jeffrey and Yu, Dian and Du, Nan and Shafran, Izhak and Narasimhan, Karthik R and Cao, Yuan , year =

  65. [65]

    2025 , booktitle=

    Multi-agent Architecture Search via Agentic Supernet , author=. 2025 , booktitle=

  66. [66]

    Cut the Crap: An Economical Communication Pipeline for LLM-based Multi-Agent Systems , year =

    Zhang, Guibin and Yue, Yanwei and Li, Zhixun and Yun, Sukwon and Wan, Guancheng and Wang, Kun and Cheng, Dawei and Yu, Jeffrey Xu and Chen, Tianlong , booktitle=. Cut the Crap: An Economical Communication Pipeline for LLM-based Multi-Agent Systems , year =

  67. [67]

    MARFT: Multi-Agent Reinforcement Fine-Tuning

    Marft: Multi-agent reinforcement fine-tuning , author=. arXiv preprint arXiv:2504.16129 , year=

  68. [68]

    Findings of the Association for Computational Linguistics , pages =

    Weize Chen and Jiarui Yuan and Chen Qian and Cheng Yang and Zhiyuan Liu and Maosong Sun , title =. Findings of the Association for Computational Linguistics , pages =

  69. [69]

    PromptWizard: Task-Aware Prompt Optimization Framework, October 2024

    PromptWizard: Task-aware prompt optimization framework , author=. arXiv preprint arXiv:2405.18369 , year=

  70. [70]

    arXiv preprint arXiv:2504.03723 , year=

    Vflow: Discovering optimal agentic workflows for verilog generation , author=. arXiv preprint arXiv:2504.03723 , year=

  71. [71]

    Motwani, Sumeet Ramesh and Smith, Chandler and Das, Rocktim Jyoti and Rafailov, Rafael and Laptev, Ivan and Torr, Philip HS and Pizzati, Fabio and Clark, Ronald and de Witt, Christian Schroeder , journal=

  72. [72]

    Multiagent finetuning: Self improvement with diverse reasoning chains.arXiv preprint arXiv:2501.05707, 2025

    Multiagent finetuning: Self improvement with diverse reasoning chains , author=. arXiv preprint arXiv:2501.05707 , year=

  73. [73]

    Darwin Godel Machine: Open-Ended Evolution of Self-Improving Agents

    Darwin Godel Machine: Open-Ended Evolution of Self-Improving Agents , author=. arXiv preprint arXiv:2505.22954 , year=

  74. [74]

    2025 , eprint=

    Huxley-Godel Machine: Human-Level Coding Agent Development by an Approximation of the Optimal Self-Improving Machine , author=. 2025 , eprint=

  75. [75]

    Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing , pages =

    Hanwei Xu and Yujun Chen and Yulun Du and Nan Shao and Yanggang Wang and Haiyu Li and Zhilin Yang , title =. Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing , pages =

  76. [76]

    Proceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics , pages =

    Archiki Prasad and Peter Hase and Xiang Zhou and Mohit Bansal , editor =. Proceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics , pages =

  77. [77]

    Findings of the Association for Computational Linguistics , pages =

    Rui Pan and Shuo Xing and Shizhe Diao and Wenhe Sun and Xiang Liu and Kashun Shum and Jipeng Zhang and Renjie Pi and Tong Zhang , title =. Findings of the Association for Computational Linguistics , pages =

  78. [78]

    Automatic Engineering of Long Prompts , booktitle =

    Cho. Automatic Engineering of Long Prompts , booktitle =

  79. [79]

    Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies , pages =

    Yao Lu and Jiayi Wang and Raphael Tang and Sebastian Riedel and Pontus Stenetorp , title =. Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies , pages =

  80. [80]

    Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing , pages =

    Yongchao Chen and Jacob Arkin and Yilun Hao and Yang Zhang and Nicholas Roy and Chuchu Fan , title =. Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing , pages =

Showing first 80 references.