Back to feed
AR
arXiv CS.AI
5/12/2026
MemQ: Integrating Q-Learning into Self-Evolving Memory Agents over Provenance DAGs

MemQ: Integrating Q-Learning into Self-Evolving Memory Agents over Provenance DAGs

Short summary

MemQ applies reinforcement learning's TD(λ) eligibility traces to episodic memory in LLM agents, propagating credit through provenance DAGs that track how memories enable future memories. Tested on six tasks, it achieves the highest success rates, with largest gains (+5.7pp) on multi-step problems requiring deep memory chains. Code will be released soon.

  • •Novel method (MemQ) for credit propagation through memory dependencies using TD(λ) traces
  • •Outperforms baselines across 6 benchmarks including code generation, OS interaction, and QA
  • •Largest improvements (+5.7pp) on multi-step tasks with deep provenance chains

Generated with AI, which can make mistakes.

#ai-agents#research-breakthrough
Read full article at arXiv CS.AI

Is this a good recommendation for you?

Explore more

From Storage to Experience: A Survey on the Evolution of LLM Agent Memory Mechanisms
arXiv CS.AI
CoCoDA: Co-evolving Compositional DAG for Tool-Augmented Agents
arXiv CS.AI
CASCADE: Case-Based Continual Adaptation for Large Language Models During Deployment
arXiv CS.AI
SkillLens: Adaptive Multi-Granularity Skill Reuse for Cost-Efficient LLM Agents
arXiv CS.AI