Scholar Feed

Agent / long-term memory

MemGPT

MemGPT: Towards LLMs as Operating Systems

Superseded baseline#6 of 63 most-superseded · first seen Oct 12, 2023

Superseded — cited as a baseline and beaten by newer methods

4 papers critique it · 3 beat it on benchmarks

What papers say

Verbatim critique sentences, each from a paper that cites MemGPT as a baseline.

However, these approaches face significant limitations in handling diverse real-world tasks. While they can provide basic memory functionality, their operations are typically constrained by predefined structures and fixed workflows. These constraints stem from their reliance on rigid operational patterns, particularly in memory writing and retrieval processes. Such inflexibility leads to poor generalization in new environments and limited effectiveness in long-term interactions.
— A-MEM: Agentic Memory for LLM Agents
“they operate in an ``open loop'' without feedback on whether the constructed memories benefit downstream tasks”
— MemBuilder: Reinforcing LLMs for Long-Term Memory Construction via Attributed Dense Rewards
“While simple to implement, these unstructured memory designs fall short when critical information is dispersed across multiple entries.”
— CAM: A Constructivist View of Agentic Memory for LLM-Based Reading Comprehension
“Early memory mechanisms in LLM-based agents typically relied on heuristic-based static workflows.”
— AtomMem : Learnable Dynamic Agentic Memory with Atomic Memory Operation

Beaten on benchmarks

Head-to-head results where a newer method reports beating MemGPT. Values are copied from the source paper's tables — verify against the cited paper.

What to use instead

Recent methods in the same sub-problem, not yet superseded in the knowledge base — arXiv benchmark leaders, not vetted production recommendations.