Continual Learning

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.

PROACTAGENT combines Experience-Enhanced Online Evolution (EXPONEVO), a structured EXPERIENCE BASE, and Proactive Reinforcement Learning-based Retrieval (PROACTRL) to jointly evolve memory and policy with retrieval as an explicit action. On SciWorld, PROACTAGENT reaches 73.50% SR versus 55.50% for GRPO+Reflexion, while cutting interaction rounds from 27.52 to 18.38.

AutoSkill composes prompt-driven modules Query Rewriting, Hybrid Skill Retrieval, Skill Extraction, Skill Management Decision, and Versioned Skill Merging to externalize user preferences and workflows into SKILL.md artifacts stored in a SkillBank. AutoSkill builds a multilingual SkillBank with 1,858 skills across four WildChat-1M subsets, showing that explicit skill artifacts can be continuously refined (e.g., professional_text_rewrite at version 0.1.34) without any parameter updates.

From Storage to Experience formalizes LLM agent memory into Storage, Reflection, and Experience, detailing mechanisms like Linear, Vector, Structured storage and Introspection, Environment, Coordination reflection. From Storage to Experience’s main result is a unified evolutionary framework that connects fragmented operating system style and cognitive science inspired memory designs into a single roadmap.

SSGM interposes a Governance Middleware, Read Filtering Gate, Write Validation Gate, and a dual substrate of Mutable Active Graph plus Immutable Episodic Log between agents and memory. SSGM unifies evolving-memory systems into a four-dimensional failure taxonomy and proves that periodic reconciliation can bound semantic drift over infinite horizons.

SleepGate augments transformers with a Conflict-Aware Temporal Tagger, Forgetting Gate, Consolidation Module, and Sleep Trigger that periodically rewrite the KV cache during sleep micro-cycles. On the PI-LLM benchmark, SleepGate achieves 99.5% retrieval accuracy at PI depth 5 and 97.0% at depth 10, while full KV cache, sliding window, H2O, StreamingLLM, and a decay-only ablation all stay below 18% across all depths.

MemEvoBench combines Misleading Memory Injection, Noisy Tool Returns, Biased User Feedback, and a Memory Modification Tool (+ModTool) to stress-test long-term memory safety in LLM agents across 7 domains and 36 risk types. On the QA Style benchmark, MemEvoBench shows Gemini-2.5-Pro’s ASR drops from 67.0% (Vanilla) to 19.0% with +ModTool in Round 1, while biased feedback can push GPT-5’s QA ASR from 59.0% to 78.0% by Round 3.

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%.

MEM1 trains an RL agent that updates a compact Internal State (IS) using Masked Trajectory for Policy Optimization, 2D Attention Mask, and Multi-Objective Task Design to jointly consolidate memory and reason. On 16-objective multi-hop QA, MEM1-7B reaches EM 1.97 vs 0.567 for Qwen2.5-14B-Instruct while using 10.4×10² peak tokens vs 38.4×10².

MemEvolve decomposes agent memory into Encode, Store, Retrieve, and Manage modules and meta evolves these components via a dual evolution process over candidate architectures. On xBench DeepSearch, MemEvolve with GPT 5 mini raises Flash Searcher pass@1 from 69.0 to 74.0 and WebWalkerQA accuracy from 58.82 to 61.18 while keeping API cost near 0.141 per query.

MemoryBench orchestrates Task Provider, User Simulator, and Performance Monitor to feed heterogeneous tasks, simulate explicit and implicit feedback, and score LLM systems across declarative and procedural memory. MemoryBench’s main finding is that state-of-the-art memory systems like A-Mem, Mem0, and MemoryOS often fail to beat naive BM25 or embedding-based RAG on partitions such as SiLo and LiLo.

Memory Decoder combines a Pre-training stage that aligns with kNN-LM distributions and an Inference interpolation mechanism that mixes Memory Decoder and base LLM outputs without changing base parameters. On Wikitext-103, Memory Decoder with 124M parameters reaches 13.36 perplexity on GPT2-small versus 14.76 for DAPT, and on specialized domains a single 0.5B Memory Decoder reduces average perplexity from 14.88 to 4.05 on Qwen2-0.5B.

RGMem builds a multi-scale memory state using Microscopic Evidence Space DL0, Structured Knowledge Space G, and renormalization operators RK1, RK2, RK3 to evolve user profiles. On PersonaMem with GPT-4.1, RGMem reaches 74.01% Avg., beating Memory OS by 8.98 points.

HippoRAG combines Offline Indexing, Online Retrieval, a schemaless knowledge graph, retrieval encoders, and node specificity to mimic hippocampal indexing for long-term memory. On 2WikiMultiHopQA, HippoRAG with Contriever achieves 89.5 R@5 vs 68.2 for ColBERTv2, a +21.3 point gain in single-step retrieval.

Long Term Memory equips OMNE with a Data Framework for LTM, Development Framework for LTM, and multi-agent collaboration over personalized memories to support AI self-evolution. On the GAIA benchmark, OMNE reaches first place, showing that LTM-driven multi-agent personalization can solve complex real-world tasks better than prior agent systems.

MD-DETR augments Deformable-DETR with frozen encoder–decoder, task-partitioned memory modules, localized query retrieval, and background thresholding to mitigate background relegation. On MS-COCO continual detection, MD-DETR reaches 50.2 mAP@A on Task 4, +11.0 over CL-DETR and +14.8 over ERD.