All AI memory papers

TRUSTMEM combines a Memory Transition Verifier, Memory Executor, Policy Model, Reference Model, and Retriever to score and optimize each memory transition for coverage, preservation, and faithfulness. TRUSTMEM achieves 65.7 average on MemoryAgentBench, improving Mem-α’s 59.2 by 6.5 points while reducing omission, corruption, and hallucination rates by up to 79.1%.

Aegis combines a four-stage content screening pipeline with provenance-weighted ranking over a memory store, using trust labels and semantic similarity to manage agent memory. Under a weak poisoning attack on LongMemEval_S, Aegis drops from 0.850 to 0.300 accuracy undefended, and even the shipped provenance weight (wt=0.15) is statistically indistinguishable from no defense (p=0.80).

Weighted Memory Tree organizes agent history into a Hierarchical Memory Tree, Dynamic Retention Scoring, Memory Controller and Lifecycle Operations, and a Utility-Aware Prompt Synthesizer that jointly decide which memories stay active. On GAIA-Text, Weighted Memory Tree reaches 33.86% accuracy with Qwen3-8B versus 20.47% for Linear History, while reducing prompt-token usage from 43.67M to 32.48M.

SP-Mem uses Privacy-Aware Memory Writing, Partitioned Storage with Privacy Mapping, and Privacy-Aware Query-Time Reasoning and Authorized Retrieval over a hybrid vector plus graph memory layer to separate sanitized facts from exact private values. On the privacy-aware benchmark, SP-Mem reaches 0.996 privacy-entity identification accuracy and reduces unnecessary privacy usage on preference-only tasks from 16.00% for Full-context to 0.33%.

WhenLoss evaluates long-context memory systems by running a fixed reader under Truncated Full Context, Oracle Evidence, Complete Stored Memory, and Retrieved Memory to expose write and retrieval bottlenecks. On 500 LongMemEval questions, WhenLoss with Expected Predictive Compression reaches 0.49 CSM Contains Match vs 0.44 for Summary LLM, reducing the write-side gap to 0.04.

Dependency-aware Semantic Garbage Collection (DSGC) augments similarity-only retention with one-hop propagation over a supplied prerequisite graph E, using semantic relevance scores ri, propagation term πi, and a greedy retained subset S under budget. On the fixed benchmark with structurally indirect prerequisites, DSGC lifts full-chain retention from 0.03 to 0.90 under the lexical encoder and from 0.23 to 1.00 under the sentence encoder, compared to the similarity-only baseline.

WorldMemArena evaluates multimodal agent memory via an Action World Interaction Loop and a four-stage lifecycle of Observe to Write, Update and Consolidate, Retrieve for Decision, and Use and Act across 461 multi-session tasks. On WorldMemArena, systems like M2A reach 86.83% memory Recall but only 50.14% QA-C, revealing that storing more correct memories does not guarantee better performance compared to baselines like UniversalRAG at 39.62% QA-C.

Zombie Agents formalizes a two-phase Infection and Trigger pipeline over Sliding Window, Retrieval-Augmented Memory (RAG), and Memory Evolution Mechanisms to implant and persist malicious payloads in long-term memory. Zombie Agents achieves cumulative Attack Success Rates near 100% in sliding-window settings and ~77% under Raw History evolution while standard indirect prompt injection baselines decay to near 0 after context eviction.

ACON combines History Compression, Observation Compression, Compression Guideline Optimization, and Compressor Distillation to rewrite agent histories and observations into concise, task-aware summaries. On AppWorld, ACON UTCO with gpt-4.1 achieves 56.5% accuracy with 7.33k peak tokens, versus 56.0% accuracy with 9.93k peak tokens for No compression.

AgentSafe secures LLM-based multi-agent systems by combining ThreatSieve for permissioned communication and identity checks with HierarCache for hierarchical memory and junk isolation. On RIOH and WCEI, AgentSafe achieves up to 96.88% defense rate in memory-based attacks and maintains CSR above 0.65, compared to baselines dropping below 0.4 without AgentSafe.

A-MEM organizes agent memory via Note Construction, Link Generation, Memory Evolution, and Retrieve Relative Memory to build an evolving, interconnected note graph. On the LoCoMo dataset, A-MEM with GPT-4o-mini reaches 27.02 F1 on Multi Hop questions, +17.87 over ReadAgent, while cutting average token length from 16,910 to 2,520.

A-MemGuard combines consensus-based validation, dual-memory structure, lesson memory, and path divergence scoring to sanitize retrieved memories and revise actions using past failures. On EHRAgent under AgentPoison, A-MemGuard reduces ASR-r from 100.0% to 2.13% and ASR-t from 100.0% to 6.38%, far below LLM Auditor and Distil Classifier.

LIGHT augments LLMs with Retrieval from the Conversation, Scratchpad Formation and Utilization, and a Working Memory buffer plus noise filtering to answer BEAM’s long-context probing questions. On the BEAM benchmark, LIGHT raises GPT-4.1-nano’s average score at 10M-token conversations from 0.109 to 0.226, a +107.3% gain over the vanilla long-context baseline.

DAM decomposes agent memory into a Read Policy, Write Policy with Sub-policy, Value Function, Uncertainty Estimator, and an Aggregate Policy that arbitrates storage operations. DAM’s main result is a formal decision-theoretic framework that clarifies the limits of heuristic memory and unifies existing memory-augmented agents under a single uncertainty-aware control perspective.

CAM builds hierarchical schemata using an incremental overlapping clustering algorithm, ego centric disentanglement, and a Prune and Grow associative strategy for retrieval. On NovelQA, CAM achieves 52.3 ACC-L versus RAPTOR’s 47.8, a +4.5 point gain while also improving efficiency in long-text reading comprehension.

CarMem manages in-car user preferences via Preference Extraction, Preference Maintenance, and Preference Retrieval over a predefined category hierarchy. On the CARMEM dataset, CarMem reaches detail-level F1 of 0.78 and optimal retrieval accuracy of 0.87 using enriched embeddings.

CodeMem reframes agents as workflow architects that combine search_functions, write_todos, execute_code, and register_skill to build deterministic, reusable Python skills. On a 25-task benchmark, CodeMem with Gemini 3 Full achieves 96% minimum correctness with 7.00 calls and 2.02M tokens, compared to 68% correctness and 2.80 calls for GPT-5 Chat.

CAT organizes context into a structured workspace with Fixed Segment, Long-Term Memory Segment, and High-Fidelity Working Memory Segment, and trains SWE-Compressor via the CAT-GENERATOR trajectory pipeline to call a context tool as a first-class action. On SWE-Bench-Verified, SWE-Compressor achieves 57.6% Pass@1 using CAT, beating the Threshold-Compression Agent at 53.8% and ReAct Agent at 49.8% under the same 32B backbone and OpenHands scaffold.

Echo combines the Multi-Agent Data Generation Framework (MADGF), EM-Train, EM-Test, and a modified user–time–assistant training paradigm to give ChatGLM3-6B temporal episodic memory. Echo scores 6.7 on easy EM-Test questions, beating GPT-4’s 5.8, and reaches 84.0 similarity vs GPT-4’s 72.3 on the same benchmark.

Evaluating Long-Term Memory for Long-Context Question Answering compares Full Context, RAG, A-Mem, RAG+PromptOpt, and RAG+EpMem memory components across semantic, episodic, and procedural memory for long conversational QA. On LoCoMo, RAG+EpMem reaches an average F1 ranking of 1.83 for Llama 3.2-3B Instruct and 1.80 for GPT-4o mini while using around 1,000 tokens per query versus over 23,000 for Full Context.

MemoryAgentBench standardizes multi-turn datasets into chunked conversations with memorization prompts, then evaluates long-context agents, RAG agents, and agentic memory agents across Accurate Retrieval, Test-Time Learning, Long-Range Understanding, and Selective Forgetting. On the overall score in Table 3, the GPT-4.1-mini long-context agent reaches 71.8 on Accurate Retrieval tasks compared to 49.2 for the GPT-4o-mini long-context baseline.

Evo-Memory restructures datasets into streaming task trajectories and plugs in components like ExpRAG, ReMem, Search, Synthesis, and Evolve to benchmark self-evolving memory in LLM agents. On multi-turn embodied reasoning benchmarks such as BabyAI and PDDL with Claude 3.7 Sonnet, Evo-Memory with ReMem reaches 0.92/0.96 and 0.83/0.95 success/progress, improving over History and ReAct baselines.

MaRS organizes agent memory into episodic, semantic, social, and task nodes with provenance, scored by a privacy-aware retention controller and governed by FIFO, LRU, Priority Decay, Reflection-Summary, Random-Drop, and Hybrid policies. On the FiFA benchmark, the Hybrid policy in MaRS achieves a composite score of ≈0.911 across 300 runs and five memory budgets, outperforming simpler policies while preserving privacy and cost efficiency.

From Human Memory to AI Memory maps human memory categories onto AI memory using the 3D-8Q taxonomy with Personal Memory, System Memory, and the Three-Dimensional Eight-Quadrant Memory Taxonomy. The main result is that From Human Memory to AI Memory systematically organizes memory in LLM-driven AI systems across eight quadrants defined by object, form, and time, connecting them to human memory types.