latent-memory

ThinkFlow is a novel end-to-end latent memory framework for lifelong conversational agents that uses probabilistic vectors to overcome textual memory bottlenecks. It enables autonomous personalization through self-evolution and test-time learning, outperforming existing memory systems.

Lngram v2 introduces an efficient and scalable latent conditional memory mechanism for transformers, decoupling memory capacity from backbone width and demonstrating consistent improvements in vision-language models through discrete addresses and structured representations.

LatentStream introduces a progressive latent working memory framework that internalizes streaming visual evidence for continuous reasoning, achieving state-of-the-art results on video benchmarks.

FocusMem introduces a latent memory interface for GUI agents that separates content retention, state-conditioned readout, and a trust gate to improve memory reliability. It consistently outperforms fixed-memory baselines across five GUI-agent benchmarks.

Proposes MemDefrag, a training-free framework that uses a middle-layer tracing signal to defragment latent memory in LLMs, achieving significantly better knowledge retention than existing methods like MemoryLLM and M+.

LaMem-VLA proposes a latent-memory-native framework that integrates short-term and long-term historical experience directly into Vision-Language-Action reasoning, enabling better performance on long-horizon robotic manipulation tasks.

EvoEmbedding is a dynamic embedding model that maintains a continuously updated latent memory to generate adaptive representations for long-context retrieval, outperforming larger specialist models and improving agentic workflows.

Latent Memory introduces a compressed representation approach for external memory in question answering, reducing token consumption and storage requirements while maintaining competitive performance across text-only and multimodal benchmarks.

ElasticMem introduces a learnable latent memory mechanism for LLM agents that adaptively allocates variable budgets to retrieved memories, improving performance on memory-intensive QA and embodied agent tasks while reducing token costs.

This paper introduces JAMEL, a framework that jointly trains agentic memory and exploration policies using novelty signals, enabling efficient exploration in open-ended environments with reduced computational costs.

This paper formulates context distillation as a latent memory management problem, proposing a framework that stores distilled contexts as independent LoRA adapters with retrieval, routing, and self-gating to improve robustness and efficiency.