episodic-memory

The paper introduces NoteBank, a training-free sparse episodic memory that augments linear attention models by archiving selected text and replaying it through native computation, boosting Gated DeltaNet's long-context retrieval accuracy on RULER from 1.7% to 70.3% at one million tokens with negligible FLOPs overhead.

A practitioner compares three memory architectures for long-running agents — full conversation history, vector-store-only retrieval, and a hybrid of structured episodic logs plus vector retrieval — concluding the hybrid approach works best and flagging memory conflicts during mid-task belief updates as a key open problem.

The paper proposes Future-Aware Recall (FAR), a framework that learns episodic memory recall for world models via future-aware predictive supervision and adaptive multi-cue scoring, automatically determining which retrieval cues (time, pose, vision, audio) to trust.

This paper introduces META, a multi-agent framework with episodic memory for financial trading that integrates specialized indicator agents to improve decision-making under market regimes.

MemBodied introduces a fixed-size episodic memory with associative state and episode anchor to enhance vision-language-action models for history-dependent manipulation tasks, achieving significant performance improvements over baselines.

CapMem is a human-annotated benchmark for episodic memory in egocentric video using captions, showing that caption-based QA outperforms direct video QA on long videos.

The article details a 25-week experiment with an AI agent's memory system, using episodic and semantic memory with nightly consolidation, avoiding LLMs in memory promotion, and employing four separate graphs for association, evidence, causation, and temporal changes, while highlighting defects like attractor basins that degrade retrieval.

MaP-WAM introduces a framework for non-Markovian robotic manipulation by separating memory-grounded planning from execution, using episodic memory to maintain fixed inference latency and achieving state-of-the-art results.

A developer shares a prototype memory architecture for multi-agent systems with separate storage for episodic, semantic, and procedural memories and intent-based retrieval, seeking feedback on scalability and design choices.

MemFuse introduces MemFuseBench, a benchmark for evaluating multi-source memory fusion in AI agents, and proposes MemFuse, a structured memory system that integrates fragmented observations while maintaining source traceability.

This paper introduces EpicStar, a framework that uses episodic memory and dynamic retrieval to help LLM agents maintain strategic coherence in long-horizon tasks like StarCraft II, achieving higher win rates with significantly fewer tokens.

This paper presents a deterministic, zero-model pipeline that compiles passively captured screen activity into structured 'activity frames' for agent memory, reducing a day of raw capture to a prompt-ready context block 86× smaller and achieving 98.4% QA accuracy. It also introduces measurements of routine overhead ratio and recurrence to model agent costs.

This paper introduces ScrubJay-MEM, an LLM agent memory system inspired by scrub jay episodic memory, which uses type-conditioned temporal decay to manage memory perishability. It also proposes the Temporal Generalization Test (TGT) benchmark and shows improved performance over existing memory systems on temporal reasoning tasks.

UniMem proposes a self-routing framework that combines episodic and parametric memory for LLM agents, enabling adaptive memory management in boundary-agnostic task streams without task labels.

This paper investigates whether long-context language models exhibit episodic-like order memory similar to humans, using a temporal order memory task based on a novel. The authors find that models show the same distance effect and uncover a single attention head reinstating temporal context, supporting temporal context reinstatement as a mechanism for episodic-like memory in LLMs.

A Twitter thread offering strategic advice on building self-learning agents that improve with use, covering dual learning sources, memory types, and data ownership.

A developer shares their experience building an AI agent with memory using the Anthropic SDK and TypeScript, explaining the differences between working, episodic, semantic, and procedural memory and the challenges of scaling memory for production.

An ex-Google engineer explains AI agent memory architecture in 12 minutes, covering working memory and three memory layers (procedural, semantic, episodic) with a summarizer to prevent token bloat, as used by Claude.

WISE proposes a long-horizon agent framework for Minecraft that enhances low-level controllers with a Causal Event Graph for episodic memory, enabling robust recall under viewpoint changes and opportunistic task reordering via causal reasoning. It also features a multi-scale progressive exploration strategy and demonstrates improved success and efficiency on long-horizon sparse tasks.

S3Mem proposes a structured spatiotemporal scene-event memory framework for long-horizon interactive question answering, using anchor-sensitive retrieval and token-budget-aware evidence interface to outperform standard RAG in multiple environments.