MemDefrag: Latent Memory Defragmentation for Large Language Models
Latent memory, which stores past knowledge fragments as per-layer hidden states, has emerged as a promising paradigm (e.g., MemoryLLM and M+) for long-term memory in large language models (LLMs). However, the paradigm suffers from significant performance degradation during memory updates, due to positional encoding misalignment and the absence of any tracing mechanism to distinguish target memory fragments from irrelevant ones. To discover such a tracing mechanism, we probe the layer-wise attention density over stored memory fragments, and find that a small set of middle transformer layers con
Lineage graph
Paper → model → repo connections mined from source citations (Tier-1 exact match).
Why these links exist
Every edge carries a method, confidence, and the source snippet that justified it — so bad links are debuggable.
- PossiblePossibly related (embedding) · 56%Anatomy of Persistent Memory's 3 Layers: Comparing ContextNest, Mem0 and Zep →
- PossiblePossibly related (embedding) · 54%las7/memharness →
- PossiblePossibly related (embedding) · 54%MemTensor/MemOS →
- PossiblePossibly related (embedding) · 53%TRACE: open-source hierarchical memory for LLM agents, 82.5% on MemoryAgentBench’s EventQA using gpt-oss-20B [P] →
- PossiblePossibly related (embedding) · 53%Breakthrough in long-context efficiency announced →
- LinkedLinked via arxiv author · 85%Ruiyi Yan →
“MemDefrag: Latent Memory Defragmentation for Large Language Models”
- LinkedLinked via arxiv author · 85%Zhuoyuan Mao →
“MemDefrag: Latent Memory Defragmentation for Large Language Models”