Tag
This paper introduces CAMeR, a memory retention framework for LLM agents that combines keyword-gated hybrid activation with adaptive weight dynamics, and presents CAMeR-Bench, a benchmark for evaluating adaptive memory retention. Experiments show that hybrid symbolic-neural gating improves retention gaps and retrieval efficiency compared to embedding-only or time-driven baselines.
This paper formulates memory retention for long-horizon language agents as a constrained stochastic optimization problem, introducing OSL-MR, a framework that enforces observability-safe learning with a Mixed-Score heuristic. Experiments show consistent improvements over existing heuristic baselines under tight memory budgets.
The paper introduces Self-Consolidating Language Models (SCoL), a framework that uses meta-reinforcement learning to write current context into model weights for continual knowledge incorporation. It demonstrates improved acquisition and retention over baselines in both QA and long-context consolidation tasks.