Entity · benchmark

Multi-hop Graph Retrieval

benchmarkactivemulti-hop-graph-retrieval-aa54b3a3·1 events·first seen May 26, 2026

Aliases: Multi-hop Graph Retrieval

Co-occurring entities

Transformers Key-Value Cache Sleep Consolidation Mechanism Cellular Automata state space model Fast Weights

More like this (12)

Multi-hop Question Answering multi-hop reasoning GLM-RAG: Graph Language Models for Graph-Based Retrieval-Augmented Generation Graph Neural Network leveraging Higher-order Class Label Connectivity for Heterophilous Graphs Knowledge Graph Random Walk heterogeneous graph memory Modern Hopfield Network dual-graph framework Graph Structure Learning Hierarchical Graph Memory Dual-Path LLM Reasoning for Multimodal Few-Shot Knowledge Graph Completion hybrid dense-sparse retrieval

Recent events (1)

6arXiv · cs.CL·May 26, 2026·source ↗

Language Models Need Sleep: Periodic Context Consolidation via Fast Weights and SSM Blocks

This paper proposes a sleep-like consolidation mechanism for transformer-based LLMs to address the quadratic scaling of attention with context length. During 'sleep' phases, the model performs N offline recurrent passes over accumulated context, updating fast weights in state-space model (SSM) blocks via a learned local rule, then clears the KV cache. The approach is evaluated on synthetic tasks (cellular automata, multi-hop graph retrieval) and math reasoning, where standard transformers and SSM-attention hybrids fail, with performance scaling with sleep duration N.

Long Context Evolution Frontier Model Releases Transformers Key-Value Cache Sleep Consolidation Mechanism +6 more