Tag
A novel memory retrieval system inspired by episodic memory theory achieves state-of-the-art 96.4% top-50 accuracy on the LongMemEval benchmark using Gemini Flash, outperforming larger Pro-based baselines by isolating retrieval quality from model capability.
This paper shows that continuously consolidating past experiences into textual memory using LLMs degrades memory utility over time, and that preserving raw episodic trajectories outperforms forced consolidation, with implications for robust agentic memory systems.
A study finds that continuously updating consolidated memories in LLM-based agentic systems degrades performance, and that retaining raw episodic trajectories is more reliable. Experiments on ARC-AGI show that even GPT-5.4 fails more often after consolidation.
The paper introduces MemQ, a method that integrates Q-learning into self-evolving memory agents by using eligibility traces over provenance DAGs to solve credit assignment problems in episodic memory retrieval.
This research demonstrates that continuously updating LLM agent memories through distillation and consolidation loops causes performance regression, even when trained on ground-truth solutions. The study finds that episodic-only retention outperforms text-based consolidation, highlighting significant flaws in current self-improvement paradigms.
This paper introduces CASCADE, a framework for deployment-time learning that allows Large Language Models to adapt continuously through episodic memory and contextual bandit optimization without modifying model parameters.