AI & Agent Memory Research, Explained Simply

Σ-Mem maintains historical competence evidence matrices M_p, a peer relationship matrix G, a memory readout r_{p,t}, and residual steering δ_{p,t} to guide multi-agent trust decisions. On the mixed counterfactual benchmark, Σ-Mem raises Qwen3-0.6B accuracy at CF@90 from 46.22% with the base model to 71.10% with Σ w/ G.

TAG routes low-confidence steps to uncertainty-based routing, filters them with guarded acceptance with rollback, chooses between bank selection across rule and exemplar memory, and prunes via evidence-based retirement inside a unified control loop. On SVAMP and ASDiv, TAG reaches 81.0% and 85.2% accuracy, improving over the 74.0% and 77.5% no-memory baselines while a compute-matched Retry baseline stays flat.

Focus Agent adds start_focus, complete_focus, a persistent Knowledge block, and an optimized Persistent Bash plus String-Replace Editor scaffold to actively compress context during long software-engineering tasks. On five hard SWE-bench Lite instances against a Baseline ReAct agent, Focus Agent achieves 22.7% token reduction (14.9M → 11.5M) while matching 3/5 = 60% task success.

ActMem transforms dialogue history into atomic facts via Memory Fact Extraction, groups them with Fact Clustering, links them through a Memory KG Construction module, and uses Counterfactual-based Retrieval and Reasoning for action-aware answers. On ActMemEval, ActMem reaches 76.52% QA accuracy with DeepSeek-V3, beating LightMem’s 63.97% by 12.55 points and NaiveRAG’s 61.54%.

ADAM combines Anchor extraction, Distribution estimation, Anchor selection, and Query generation to adaptively probe agent memory via an auxiliary generator and entropy based selection. On the EHRAgent benchmark with Llama2-7b-chat, ADAM reaches EQ=77 and ASR=1.00, compared to MEXTRA’s EQ=44 and ASR=0.89.

AdaMem organizes dialogue history into Working Memory, Episodic Memory, Persona Memory, and Graph Memory coordinated by a Memory Agent, Research Agent, and Working Agent. On LoCoMo with GPT-4.1-mini, AdaMem achieves 44.65 F1 overall, beating the best baseline LangMem at 41.76 F1 by +2.89.

AdaMEM combines Long-Term Trajectory Memory, Short-Term Strategy Memory, ADAMEM-HIGH, ADAMEM-LOW, and STEP-MFT to adapt agent behavior at each decision step without parameter updates. On ALFWorld unseen, AdaMEM-LOW achieves 58.2% Success Rate versus 52.2% for Synapse, and on WebShop reaches 74.2 Task Score versus 71.4 for the No Memory baseline.

A-MAC scores candidate memories using Utility, Confidence, Novelty, Recency, and Type Prior combined by a learned linear admission policy with Algorithm 1 A-MAC Memory Admission. On the LoCoMo benchmark, A-MAC achieves F1 0.583 and 2644 ms latency, improving F1 by 0.042 and reducing latency by 1187 ms compared to A-mem.

LingBot-World combines a Data Engine, Fundamental World Model, Action-Conditioned World Model, and Post-Training causal adaptation to turn a 28B-parameter video generator into a real-time interactive world simulator. On the VBench benchmark, LingBot-World achieves a dynamic degree of 0.8857 versus 0.7612 for Yume-1.5, while also improving imaging quality to 0.6683.

Adaptive RAG Memory (ARM) augments a standard retriever–generator stack with a Dynamic Embedding Layer and Remembrance Engine that track usage statistics and apply selective remembrance and decay to embeddings. On a lightweight retrieval benchmark, ARM achieves NDCG@5 ≈ 0.9401 and Recall@5 = 1.000 with 22M parameters, matching larger baselines like gte-small while providing the best efficiency among ultra-efficient models.

Aeon restructures LLM memory using the Atlas, Trace, Semantic Lookaside Buffer, Write Ahead Log, and Sidecar Blob Arena inside a zero copy Core Shell kernel. Aeon achieves 4.70 ns INT8 dot products, 3.09 µs Atlas traversal at 100K nodes, 3.1× compression, and P99 read latency of 750 ns under 16 thread contention compared to FP32 and flat scan baselines.

AGENTCL evaluates continual learning in language agents by contrasting naive and compositional task streams with a two-pass protocol and MEMPROBE’s interaction memory, insight memory, and skill memory. On the CodeEval-Pro compositional stream, AGENTCL shows MEMPROBE achieving 66.7% first-pass accuracy versus the memoryless ReAct’s 48.8% (+17.9 pp), while naive streams show much smaller separations.

AgenticAI-DialogGen chains ChatPreprocessor, KnowledgeExtractor, TopicAnalyzer, KnowledgeGraphBuilder, PersonaGenerator, DuelingChat Agent, ConversationValidator, ConversationRefiner, QAGeneration, and PostProcessing to turn raw multi-session chats into topic-guided, persona-grounded conversations with explicit short- and long-term memories. On the TGC / KG memory QA benchmark, Mistral-7B fine-tuned within AgenticAI-DialogGen achieves 87.36 F1, compared to GPT-4’s 83.77 F1 in a zero-shot setting on the same task.

Agentic Memory (AgeMem) exposes memory management tools, a three-stage progressive RL strategy, and step-wise GRPO directly inside the agent policy to jointly control long-term and short-term memory. On Qwen3-4B-Instruct, AgeMem attains 54.31% average performance across ALFWorld, SciWorld, PDDL, BabyAI, and HotpotQA, exceeding the best baseline A-Mem at 45.74%.

AgenticSTS composes each decision from L1 protocol instructions, L2 state-typed prompts, L3 game knowledge, L4 episodic memory, and L5 skill library instead of an accumulating transcript. On fixed A0 Slay the Spire 2 runs, AgenticSTS with L5 skills wins 6/10 games versus 3/10 for the no-scaffold baseline, and climbs to A6–A8 in auto-mode streams.

AgentIR combines a SIMD-Accelerated BM25 engine, a Temporal Partitioned Index, an Agent-Aware Fusion layer, and a Cascade Router over a shared CSR substrate to adapt retrieval per query. On LongMemEval, AgentIR’s confidence-triggered cascade skips dense on 63% of queries for a 2.67× latency reduction at parity LLM-judged accuracy, and on BEIR AgentIR’s BM25 core is up to 29× faster than Pyserini 8T at matching nDCG@10.

AgentKVShift reuses KV cache per retrieved memory unit using Probe Selection, Mean Estimation, and AgentKVShift Correction to adjust stale keys and values. On LoCoMo with AMem and Qwen2.5-3B-Instruct, AgentKVShift achieves 0.319 F1 at recompute ratio 0.1, compared to 0.178 for CacheBlend and 0.125 for ProphetKV.

AgentMemBench wires five strategies—In-Context Windowing (ICW), External Key-Value Store (EKV), Graph-Based Episodic Memory (GEM), Compression-Based Summarisation (CBS), and Web-Augmented Memory (WAM)—into a shared Store Read Generate interface over multi-session dialogues. AgentMemBench’s main result is that EKV achieves macro Recall@5 0.792 on LoCoMo, MultiDoc2Dial, and MSC, a +0.324 gain over ICW’s 0.468, while also leading MRR, Answer F1, and Faithfulness.

Agent Memory Below the Prompt stores each agent’s KV state in a block pool, quantizes it via a Q4 pipeline, reloads it with BatchQuantizedKVCache, and reuses it across phases using cross-phase context injection. On Gemma 3 12B, Agent Memory Below the Prompt reduces cold TTFT from 172,096 ms to 1,264 ms at 32K context (136×) compared to FP16 prefix caching baselines like vllm-mlx.

Agent Memory decomposes agent workloads into ingestion, memory construction, storage, retrieval, prompt assembly, generation, and maintenance, and classifies ten systems across four paradigms. Agent Memory’s profiling on MemoryAgentBench and MemoryArena reveals over 47× spread in lifecycle energy per correct answer and two orders of magnitude differences in serving latency across paradigms.

Agent Memory Distillation builds Workflow memory, Subtask memory, and Function memory from successful teacher trajectories and injects them via proactive and reactive retrieval to guide small agents. On AppWorld, Agent Memory Distillation lifts Qwen3-4B from 14.88% to 49.40% accuracy (+34.52%p) compared to the zero-shot baseline.

EAL-BENCH wires a memory writer, executor, canonical ledger, and authorization predicate together to isolate how agent memory encodes and propagates permissions. Across procurement, cybersecurity, and finance, typed incremental memory forms false authority for up to 50.2% of unauthorized requests, and executors act on it in 98.6% of matched trials.

AgentSM builds a planner agent, schema linking agent, trajectory store, and composite tools that turn past execution traces into reusable, structured semantic memory. On Spider 2.0 Lite, AgentSM with Claude 4 Sonnet achieves 44.8% execution accuracy versus 28.7% for SpiderAgent, while shortening trajectories and latency.

AGENTSYS organizes a Main Agent, Worker Agents, Intent Schemas, and an Alignment Validator into a hierarchical memory system that isolates raw tool outputs and only admits schema-validated JSON. On AgentDojo, AGENTSYS reaches 52.87% attacked utility and 0.78% ASR versus 48.27% and 30.66% for the No Defense baseline.