Scholar Feed

Agent / long-term memory

MemSkill

MemSkill: Learning and Evolving Memory Skills for Self-Evolving Agents

Superseded baseline#19 of 63 most-superseded · first seen Feb 2, 2026

Superseded — cited as a baseline and beaten by newer methods

2 papers critique it · 1 beat it on benchmarks

What papers say

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

However, most existing agent-memory approaches still rely on unweighted or weakly weighted relations, where an edge primarily indicates the existence of a connection rather than its query-dependent utility.
— HAGE: Harnessing Agentic Memory via RL-Driven Weighted Graph Evolution
“These approaches use smaller ML models rather than full LLMs for structuring, but still require task-specific training data and GPU inference during ingestion or search.”
— SmartSearch: How Ranking Beats Structure for Conversational Memory Retrieval

Beaten on benchmarks

Head-to-head results where a newer method reports beating MemSkill. 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Recent methods in the same sub-problem, not yet superseded in the knowledge base — arXiv benchmark leaders, not vetted production recommendations.